AI News Archive: June 16, 2026 — Part 10
Sourced from 500+ daily AI sources, scored by relevance.
- Apple @ Work Podcast: Leebry aims to unify your SaaS tools for AI use
Apple @ Work is exclusively brought to you by Mosyle , the only Apple Unified Platform. Mosyle is the only solution that integrates in a single professional-grade platform all the solutions necessary to seamlessly and automatically deploy, manage & protect Apple devices at work. Over 45,000 organizations trust Mosyle to make millions of Apple devices work-ready with no effort and at an affordable cost. Request your EXTENDED TRIAL today and understand why Mosyle is everything you need to work with Apple . In this episode of Apple @ Work, Dan Jaenicke, Director of B2B Product Strategy at MacPaw, joins the show to talk about Leebry .
Score: 28🌐 MovesJun 16, 2026https://9to5mac.com/2026/06/16/apple-work-podcast-leebry-aims-to-unify-your-saas-tools-for-ai-usage/ - MiMo Code Is the Open Source Answer to Claude Code
MiMo Code Is the Open Source Answer to Claude Code DevOps.com
- Ex-Lush chief’s lawyers hike costs to ensure their AI model isn’t trained by juniors
Lawyers for the former chief executive of cosmetics giant Lush, Andrew Gerrie, have hiked their fees ahead of a trial this month, partly as they will not allow junior staff to train their AI model, City AM can reveal. A court heard this morning that Gerrie’s legal team from Brown Rudnick have pushed up a [...]
Score: 28🌐 MovesJun 16, 2026https://www.cityam.com/ex-lush-chiefs-lawyers-hike-costs-to-ensure-their-ai-model-isnt-trained-by-juniors/ - Never Let the LLM Write the Joins
The subject graph, and the one architectural line that made me comfortable putting “talk to your database” in front of real users with real permissions. Part 3 of 4 on building a conversational analytics engine. ~11 min read. Part 2 built a map: three databases introspected and enriched into a domain graph. Tables, columns, join paths, business meaning, and the security rules, all sitting in a graph with a fast in-memory snapshot in front of it. A map answers nothing on its own. Nobody asks “describe the schema.” They ask “how many orders did Bike World place last quarter, in my territory,” and they expect a number, computed from data they are actually allowed to see. This article is the machine that does that. Back in Part 1, I listed seven walls that break text-to-SQL on real data. This is where four of them fall: the fan-out trap (Wall 3), ambiguity (Wall 4), security (Wall 6), reproducibility (Wall 7), plus the cross-source half of Wall 5. If Part 2 was the “why a map,” this is the “why a particular shape of pipeline.” And I want to lead with that shape, because it is the single most important idea in this entire series. The big idea: a query is not “hand the question to an LLM and let it call some tools.” I split it in two. Phase A is deterministic planning. It decides what to do (tables, entities, filters), scoped to who is asking. One LLM call, for reading intent. Everything else is code. Phase B is LLM-driven execution. It does the plan (writes SQL, runs it, recovers, answers). The line between them is the product. Security is never the model’s job. Joins are never the model’s job. The model does the creative parts; code does the parts that must be safe and reproducible. TL;DR The subject graph resolves the specific things a user names (“Bike World” becomes StoreID 620). It is a cache, not a copy. Phase A plans deterministically and produces a frozen plan. Phase B is a bounded ReAct loop that executes it. The LLM writes only the SELECT and WHERE. Joins come from the graph. Security is injected by rewriting the SQL. Cardinality is fixed by a planner. None of those are left to the model. Honest limit: today the conversation is single-shot . Follow-up memory is built but not wired. Throughout, one example: a sales rep I will call Dana , scoped to the Southwest territory, allowed to see sales data, asks “Show orders for customer Bike World.” Watch what each stage does to that sentence. The subject graph: concepts vs. actual things The domain graph from Part 2 knows about types . It knows CUSTOMER is a concept that lives in a table with a name column. It does not know that “Bike World” is a specific store with the key StoreID 620. That is an instance, and instances live in the subject graph . A canonical entity in the subject graph has: a stable id (something like entity:organization:bikeworld_a1b2c3d4) a name and type a set of aliases a list of source links , each pointing at exactly where the entity lives: StoreID 620, in the StoreID column, of Sales.Store, in the SQL Server source an access level , so resolved entities are protected by the same machinery as tables The defining property: it is a cache, not a copy. It is not pre-loaded with every customer. That would be copying the data, the one thing I refuse to do. It starts essentially empty and fills lazily , remembering only the names real users actually ask about. The rows stay in the source database. The subject graph just remembers the mapping from “a name a human typed” to “a key the database understands.” Resolving a name (the cascade) When the pipeline needs to turn “Bike World” into a key, it runs a cascade, ordered from cheapest-and-most-certain to most-expensive-and-least-certain. It stops as soon as it is confident. entity-resolution-cascade How to read it (with the real numbers, because the numbers are the design): Cache keyed on the name. Stores unfiltered results; RBAC applies after, so users with different permissions can share an entry. Exact match: canonical name scores 0.95 , an alias 0.90 . One clean hit above the 0.70 threshold short-circuits the whole thing. Fuzzy match: the rapidfuzz ratio scorer at a threshold of 80/100 , with tiered confidence so it never pretends to be as sure as an exact match ( 0.65 to 0.85 ). Semantic match: I will be straight with you. It is a placeholder. It sits in the right position in the cascade and returns nothing today. The plan is embedding similarity; the wiring is not done. Wherever you read “exact, fuzzy, semantic, source,” know the semantic rung is a stub. Source-DB lookup: hit the real table. An exact name scores 0.95 , and a success gets written back into the subject graph so the next person skips to a cache hit. That is the lazy cache filling. Two design choices I would defend in any review: “Ambiguous” is a first-class outcome. Two close candidates returns “did you mean the Seattle or the Dallas store” , not a coin flip. (This is the fix for Wall 4.) Resolution is fail-closed. The cascade runs and caches unfiltered, but nothing returns until it passes RBAC: can this user see the entity, its links, and any restricted properties. No scope and no trusted-caller flag means deny , not leak. Phase A: planning, stage by stage Dana’s question comes in. Phase A turns it into a plan without writing one character of SQL. User context assembly. Load who Dana is: role (sales rep), territory (Southwest), permissions (sales schema), saved vocabulary and preferences. Everything downstream is scoped by this. Intent classification (the one LLM call). Returns a structured verdict: intent type (a SQL query), entity types (CUSTOMER and the order entity), confidence, and a default join type. A second focused call pulls “Bike World” out as a value tied to CUSTOMER. A type gate then throws out any extracted value that does not fit the column’s type. Semantic resolution (the 2-pass vector search from Part 2). Turns “orders” into the order entity by meaning, not spelling , and flags ambiguity when two concepts score within a hair of each other. Scope resolution. Turns words like “my team” into row-filter hints, on pre-derived rules, in a few milliseconds. One subtlety I am careful about: this produces relevance hints, not security . Convenience and access control are different things, and conflating them is how you build something convenient and insecure. Routing (the deterministic core). Maps entity types to real tables, assigns each entity a role (the order entity is the subject, CUSTOMER is a filter) using a four-rule scoring system with no LLM , finds the foreign-key join path by traversing the domain graph, grounds “Bike World” to StoreID 620, and detects whether the query spans sources. RBAC Checkpoint 1. The first of two security gates. Coarse and early: drop any table the user may not see before the model ever sees the schema . Dana keeps her sales tables. Had she asked about HR salaries, that table would vanish here, and the model would never learn its column names. The output of Phase A is a frozen plan plus a warm entity cache. No SQL exists yet. Phase A decides everything that has to be decided safely, and hands the model a plan it would have to work hard to misuse. Two phase pipeline How to read it: the question flows top to bottom through Phase A (all blue, deterministic, except the one purple intent call), produces a frozen plan, and only then enters the Phase B loop where the model actually works. The two red boxes are the security gates, one per phase. Phase B: where the model actually works Phase B is a small state machine with three nodes: agent node: the LLM reasons and decides the next action tools node: executes exactly one tool, appends the result synthesize node: writes the final answer It loops between thinking and acting, then synthesizes. And it is bounded : at most 10 reasoning iterations and 15 tool calls per query, with a 30-second timeout per tool. Those limits are not decoration. They are the difference between “the agent retries a failed query once and succeeds” and “the agent burns your budget in an infinite loop of slightly-wrong SQL.” (This is part of the fix for Wall 7.) The agent has 11 tools but most queries touch two or three. A nice detail: resolve_entity looks like a live tool, but because Phase A already resolved Bike World into the warm cache, it usually just returns the cached answer if the confidence is above 0.5. The same operation is both "a tool the agent can call" and "something Phase A already did." Generating SQL without letting the model touch the dangerous parts This is where the two-phase philosophy pays off most concretely. Sql generate and execute How to read it: the model (the one purple box) writes only a slice of the query. Everything else is deterministic. The joins come from the graph, not the model. A join-tree builder traverses the foreign-key edges. No path between two tables means a loud error, never a guessed join. (This is the fix for Wall 2.) The model gets a whitelist of real columns and writes only the SELECT, WHERE, GROUP BY, and limit. Every column it returns is validated. A made-up or unqualified column rejects the whole generation. Unvalidated SQL never reaches the database. Filters get merged in by code, not the model. The load-bearing rule, again: the LLM never writes the joins. A made-up join is a silently wrong answer, and silently wrong is the worst failure an analytics system has. The fan-out trap (Wall 3, finally solved) Let me tell you about the bug that taught me to respect cardinality. Ask for total revenue per customer. The query joins customers to orders to line-items. The moment you SUM across a one-to-many join, the value repeats once per child row. A customer with one $100 order and five line items reports $500 . The SQL is perfect. The number is garbage. Nobody notices. So I built a deterministic planner that rewrites the SQL to prevent it. It keys entirely on the cardinality_class metadata from Part 2 (the PRESERVING vs MULTIPLYING tags), touches no table names, never calls the model, and fails open (if it cannot reason safely, it returns the original). Its strategies, picked by what the query actually references: Prune a multiplying join that is joined but never used (transitively, so chains collapse). Semi-join : a child used only in the WHERE becomes an EXISTS subquery, which filters without multiplying. Per-grain subqueries : two measures at two different grains get computed separately and joined back with a full outer join and null-safe matching. Count anchoring so a plain row count does not lie either. Why is this in deterministic code and not the prompt? Because “be careful about fan-out” is right most of the time, and “most of the time” is not a number you put on a financial dashboard. Cardinality correctness is provable from graph metadata, so I prove it. The second security gate (Wall 6) When the agent calls execute_sql, the SQL is correct but not yet safe. RBAC Checkpoint 2 runs here, and it is the fine-grained one. Where Checkpoint 1 decided which tables , Checkpoint 2 decides which rows and columns . It rewrites the SQL’s syntax tree: it injects mandatory and row-level filters straight into the WHERE. Dana's query silently gains a Southwest-territory constraint. A failed access check injects WHERE 1=0 , a silent deny that returns zero rows rather than leaking the existence of restricted data. Restricted columns are masked on the way out. The entire RBAC path is deterministic. There is no LLM anywhere in it , by design. Access control you cannot reproduce is access control you cannot defend. Then the agent synthesizes: it writes “Bike World placed 8 orders last quarter,” with any caveats (like “limited to your territory”) and a confidence, plus the SQL and rows underneath. The validation mesh Threaded through both phases is a layer I think of as a mesh, not a stage: eight validators, forty-four checks , firing at hook points all over the pipeline. Validation mesh How to read it: Three validators run in Phase A, and two of them (route, entity) can hard-abort the query before the agent loop ever starts. No point generating SQL for a query that was never going to work. The rest run in Phase B, guarding execution and the answer. Most checks are deterministic; a few are LLM calls (does the answer follow from the data), and those get a strict budget of three LLM validation calls per query . Two choices that matter: Every validator is wrapped in try/except. A validator that crashes, or is disabled, never blocks a query. Safety code that can take down the system it protects is a liability, so it is built to fail safe. Checks have severity tiers: block, warn (becomes a caveat, lowers confidence), or log-only. Not every problem should stop a query. Some just attach a footnote. Shipped vs. next (an honest naming note): the “memory validator” in this mesh has nothing to do with conversation memory. It checks whether a cached result went stale. The actual conversation is single-shot today . The read side of memory is wired (the system loads your saved vocabulary and scope on every query), but the write-back that would let Dana ask “and the quarter before that?” and have it understand “that” is built and not yet connected . Ask a follow-up today and it is treated as a brand-new question. I would rather say that than imply a multi-turn experience I have not finished. When the data lives in two databases at once (Wall 5) Some questions span sources: SQL Server sales joined to Postgres production. There is no cross-database JOIN here, and no copying one source into the other. A federated executor decomposes the query by source. It generates and runs a separate, independently RBAC-enforced query against each source, pulling only the rows that survive each source’s own rules. It joins the results in the application layer , in memory, by matching on entity keys. The requested join type (inner, left, right, full outer) is honored by deciding which side’s unmatched rows to keep. Each database only ever sees a query for its own data, scoped to what the user may see. The rows are stitched together only after each has been cleared. The data never leaves its home system. That property is non-negotiable. Why keep both a plan and a loop? Fair question: if Phase A plans everything, what is the loop for? They fail in different ways, and I wanted coverage on both. Job ( Deterministic plan (Phase A) : decide what : tables, entities, filters; ReAct loop (Phase B) : do it: write SQL, run, answer, recover) LLM? ( Deterministic plan (Phase A) : only intent classification; ReAct loop (Phase B) : yes: SQL text + prose) Gives you ( Deterministic plan (Phase A) : safety, reproducibility, auditability, speed; ReAct loop (Phase B) : execution, self-correction, tool choice) Costs you ( Deterministic plan (Phase A) : adaptability (the plan is frozen); ReAct loop (Phase B) : nondeterminism (must be fenced)) The loop earns its keep when the model generates SQL with a wrong column, the database errors, and the agent reads the error, regenerates, and succeeds. A straight pipeline cannot self-heal. The plan earns its keep by making sure the model can retry SQL all it wants but can never retry its way into a forbidden table or an invented join. The honest current limit: today the agent is locked into Phase A’s plan. It can fix a bad query, but not a bad plan. If routing picked the wrong table, retrying SQL will not help, because the agent cannot re-route mid-flight. The roadmap is to make the planning steps themselves into tools the agent can re-call inside the loop. That is the difference between a system shaped like an agent and one that is fully agentic, and it is the next real piece of work. Key takeaways Split the stochastic from the deterministic, and be ruthless about which is which. The model understands questions and writes creative SQL well. It is unaccountable at deciding access and inventing joins. Opposite sides of a hard line. Resolve the things a user names through a cheap-to-expensive cascade, and let “ambiguous” ask a question instead of guessing. A confident wrong answer is worse than an honest “which one?” Prove what you can prove with metadata instead of asking the model to be careful. Cardinality and access control are both provable from the graph, so they live in code, never in a prompt. Make your safety layer unable to take down the thing it protects. Wrap it, budget it, tier it. In Part 4 , I pick up the thread I keep deferring: documents. Everything here was structured data. Real questions often need a number and a paragraph from a PDF. The subject graph is meant to be the hub that ties the two together, and that is the most forward-looking, and most honest, story of the four. Next up, Part 4: Two Pipelines, Three Graphs, on fusing structured and unstructured retrieval with RAG. Never Let the LLM Write the Joins was originally published in Towards AI on Medium, where people are continuing the conversation by highlighting and responding to this story.
Score: 28🌐 MovesJun 16, 2026https://pub.towardsai.net/never-let-the-llm-write-the-joins-fe5d1f1ccbe0?source=rss----98111c9905da---4 - How AI-Powered CMS Platforms Are Transforming Enterprise Content Operations
For years, enterprise content management was largely a publication tool. How do you get the right content, in the right format, to the right channel, without breaking workflows that span dozens of markets and hundreds of contributors? The answer was usually a combination of manual processes, siloed systems, and large coordination teams that grew historically […] The post How AI-Powered CMS Platforms Are Transforming Enterprise Content Operations appeared first on AI News .
- Physical AI - Building Affordable & Sustainable Housing
The US faces a severe housing affordability crisis. Reframe Systems, founded by ex-Amazon engineers uses physical AI & robotics in micro-factories to solve this crisis.
Score: 28🌐 MovesJun 16, 2026https://www.forbes.com/sites/sabbirrangwala/2026/06/16/physical-aibuilding-affordable--sustainable-housing/ - Business automation platform TrackerSuite AI raises pre Series A round
Business automation platform TrackerSuite AI has raised Rs 6 crore in a pre-Series A funding round led by a UAE-based family office and advised by Bestvantage Investments, with participation from Pontaq.VC, Shubhan Ventures, Candle Advisors, and strategic angel investors. The fresh funds will be used to strengthen artificial intelligence capabilities, expand enterprise-grade offerings, accelerate customer acquisition, deepen ecosystem integrations, and support the company's global expansion plans, TrackerSuite said in a press release. Co-founded in 2021 by Neha Chandra and Rishab Chandra, TrackerSuite.AI is an AI-powered business operating system that enables small and medium-sized businesses to digitise and automate critical workflows across operations, sales, workforce management, and inventory processes. Serving over 1,800 businesses across 25 countries, the platform helps organisations improve efficiency, enhance visibility, and scale growth through intelligent automation and actionable insights. TrackerSuite offers an integrated platform that enables businesses to manage sales, workforce operations, inventory, task execution, attendance, and reporting through a single AI-powered operating system. The platform is designed to help organisations move away from fragmented workflows, manual processes, and multiple disconnected software tools. TrackerSuite.AI claims that it serves more than 1,800 businesses across 25 countries, including India, the Middle East, and Southeast Asia, helping organisations improve productivity, operational visibility, and execution efficiency through intelligent automation.
Score: 27💰 MoneyJun 16, 2026https://entrackr.com/snippets/business-automation-platform-trackersuite-ai-raises-pre-series-a-round-12043831 - South Park Commons backed Maya Research wants to build a voice interface that speaks like a local
Maya Research is developing conversational AI models designed to speak and respond like native speakers across languages and cultural contexts, aiming to serve the next four to five billion internet users. The startup, which raised $1.
- You’ve heard from my husband, Eric Schmidt, about AI. Now here’s my take
You’ve heard from my husband, Eric Schmidt, about AI. Now here’s my take San Francisco Chronicle
- This New Gadget Uses Custom Scents and AI to Improve Your Sleep
Kimba is an AI-powered sleep technology system that releases personalized scents to help you sleep better. Here’s how it works.
Score: 27🌐 MovesJun 16, 2026https://www.cnet.com/health/sleep/new-gadget-doesnt-just-track-sleep-uses-custom-scents-ai-improve-it/ - How I Continually Improve My Claude Code
I’ve written a lot of previous articles on techniques I apply when using Claude Code to get the most out of it. Continue reading on Towards AI »
Score: 27🌐 MovesJun 16, 2026https://pub.towardsai.net/how-i-continually-improve-my-claude-code-2fb58ca89ea9?source=rss----98111c9905da---4 - WhyBrilliant raises €1M to scale AI job matching, backed by Merantix
Berlin-based hiring platform WhyBrilliant has launched publicly with€1 million in pre-seed funding from Merantix. The company is developing anAI-powered recruitment platform designed to help professio...
Score: 26💰 MoneyJun 16, 2026https://tech.eu/2026/06/16/whybrilliant-raises-eur1m-to-scale-ai-job-matching-backed-by-merantix/ - Papa Johns Can Predict When Your Fridge Is Empty
Papa Johns tapped NBCUniversal, Instacart and the dentsu-owned media agency Carat for help reaching consumers when they’re low on groceries – and thus more likely to be swayed by a mouth-watering ad. The post Papa Johns Can Predict When Your Fridge Is Empty appeared first on AdExchanger .
Score: 26🌐 MovesJun 16, 2026https://www.adexchanger.com/tv/papa-johns-can-predict-when-your-fridge-is-empty/ - Deals in brief: Tin Men Capital backs Pints AI, Malaysia’s GreatAsic raises funding, latest China investments, and more
Bringing you the latest updates on funding and investment activity across the Asia Pacific.
- Poor data quality is breaking AI ambitions: Anand Ramamoorthy, Director APAC Data Governance and Quality, Informatica, Salesforce
Low data quality is hindering enterprises from advancing AI systems from pilot phases to fully operational agents
- LinkedIn Networking Feels Fake Now. Here’s What Actually Works in the Age of AI DMs
As AI-powered outreach explodes, the value of genuine, deliberate LinkedIn networking has only increased.
- From VIP suites to fraud allegations: A Delco gym manager built an AI-fueled startup around cancel culture and sports, then fumbled it away
From VIP suites to fraud allegations: A Delco gym manager built an AI-fueled startup around cancel culture and sports, then fumbled it away Inquirer.com
Score: 25🌐 MovesJun 16, 2026https://www.inquirer.com/news/lifebrand-social-media-ai-company-fraud-20260616.html - Dump all your AI subscriptions for this $70 all-in-one lifetime platform with GPT, Claude, and more
Access GPT-4o, Claude 3 Opus, Gemini Pro 1.5, Llama 3, and Mistral in one AI platform with 1min.AI’s $59.97 lifetime plan for writing, design, and productivity tools.
Score: 24🌐 MovesJun 16, 2026https://mashable.com/tech/june-16-1minai-advanced-business-plan-lifetime-subscription - The new weapons in the newsletter wars are Claude and ChatGPT
The new weapons in the newsletter wars are Claude and ChatGPT Business Insider
Score: 24🌐 MovesJun 16, 2026https://www.businessinsider.com/the-newsletter-wars-are-heading-to-claude-and-chatgpt-2026-6 - Tailscale updates Aperture to help businesses manage fast-changing AI tools
Toronto startup adds new features aimed at shadow AI, corporate data access, and agent controls. The post Tailscale updates Aperture to help businesses manage fast-changing AI tools first appeared on BetaKit .
Score: 24🌐 MovesJun 16, 2026https://betakit.com/tailscale-updates-aperture-to-help-businesses-manage-fast-changing-ai-tools/ - Bundle Your ChatGPT, Claude, and Other AI Subscriptions for 41% Off
Bundle Your ChatGPT, Claude, and Other AI Subscriptions for 41% Off PCMag
Score: 23🌐 MovesJun 16, 2026https://www.pcmag.com/deals/bundle-your-chatgpt-claude-and-other-ai-subscriptions-for-41-off - Emerging markets offer no escape from AI risk
Emerging markets offer no escape from AI risk The Telegraph
Score: 23🌐 MovesJun 16, 2026https://www.telegraph.co.uk/business/2026/06/16/emerging-markets-offer-no-escape-from-ai-risk/ - Stop copy-pasting prompts across GPT-4o, Claude, and Gemini — this tool combines them for a one-time $55
Compare outputs from GPT-4o, Claude Sonnet, Gemini, DeepSeek, Llama, and more with ChatPlayground AI’s lifetime plan for side-by-side AI testing and prompt optimization.
Score: 22🌐 MovesJun 16, 2026https://mashable.com/tech/june-16-chatplayground-ai-unlimited-plan-lifetime-subscriptions - AI Red Teaming Explained: What It Is and Why You Need It
With AI adoption accelerating, testing systems under adversarial conditions has become increasingly important. It enables organisations to identify vulnerabilities before deployment and strengthen overall system safety. Explore what AI red teaming is, why it matters and the leading companies offering AI red teaming consulting services. What Is AI Red Teaming? AI red teaming tests artificial […] The post AI Red Teaming Explained: What It Is and Why You Need It appeared first on AI News .
Score: 22🌐 MovesJun 16, 2026https://www.artificialintelligence-news.com/news/ai-red-teaming-explained-what-it-is-and-why-you-need-it/ - Siri AI Might Tell You to Take Breaks, Remind You It's Not a Real Person
Code strings discovered in iOS 27 suggest that Apple may be planning to show users a break reminder after especially long Siri AI conversations. Strings of code in the first developer beta of iOS 27 refer to a "Take a Break Message" that would remind users they have been in a conversation for an extended period and that Siri is not a real person. Based on the shared code, the reminder appears to read: "You've been in this conversation for [n] hours - consider taking a break. Siri is not a person, but will be here when you're ready to continue." Where screen time tools typically focus on usage duration, Apple appears to be specifically addressing the risk of parasocial attachment to AI, building in a prompt that explicitly reframes Siri as a tool rather than a companion. The concern is part of a broader conversation across the AI industry about unhealthy usage patterns. Both OpenAI and Google have moved to add guardrails to their chatbot products, and Anthropic has been spotted nudging Claude users toward healthier habits after long sessions. Apple touched on several privacy and responsibility considerations for Siri AI during last week's WWDC keynote , but did not address the question of extended conversations. The existence of these code strings suggests the company is thinking about the issue behind the scenes. It is not yet clear how Apple would trigger the reminder. The code does not appear to specify a fixed time threshold, suggesting the company may use conversation length in combination with other signals to determine when to display the message. Tags: Siri , Siri AI This article, " Siri AI Might Tell You to Take Breaks, Remind You It's Not a Real Person " first appeared on MacRumors.com Discuss this article in our forums
Score: 22🌐 MovesJun 16, 2026https://www.macrumors.com/2026/06/16/siri-ai-might-tell-you-to-take-breaks/ - ‘Dangerous’ AI Models Are Coming No Matter What
The US government crackdown on Anthropic’s Claude Fable 5 and Mythos 5 hides a glaring truth: AI models with advanced hacking capabilities will soon be the norm.
Score: 22🌐 MovesJun 16, 2026https://www.wired.com/story/dangerous-ai-models-are-coming-no-matter-what/ - Does preservation make sense before we know how to revive?
My name is Aurelia Song and I hope to make whole-body, human, end-of-life preservation for future revival a new global tradition. I care about it so much I've dedicated my life to it. [1] The biggest objection I get to end-of-life preservation goes like this: "We can't revive today, so we can't prove that preservation works. Therefore preservation probably doesn't work. We shouldn't bother with preservation until we can revive." I call this the immediate revival objection . I respect the immediate revival objection. If your standard of evidence is full recovery, then you don't need any knowledge of how people or mental processes work on the inside to evaluate preservation; you can just observe that they survive a round trip. I think requiring revival, now, is reasonable a priori —it's analogous to how I feel when people talk about new kinds of quantum computers: I'll believe it when they're actually doing something useful. However, in my opinion the logic of the immediate revival objection is too conservative when it comes to end-of-life preservation. Instead, I think that as a scientific community, we've known enough to preserve people for at least 30 years. I think we can and should start preserving people today. I think if you knew what I know, you'd agree. Why do I think that? What's my response to the immediate revival objection? How do I know what I think I know? The San Diego Frozen Zoo Dr. Kurt Benirschke started the San Diego Frozen Zoo in 1975. It just celebrated its 50th anniversary . Benirschke had the visionary idea to preserve cells from endangered animal species using liquid nitrogen, with the belief that people in the future would probably want them. DNA had been shown to be a durable, double-helix-shaped polymer in 1953, and successful cryopreservation of sperm had been achieved in 1949, so Dr. Benirschke was in a position to understand that animal cells each separately held the "molecule of heredity" and were preservable by cold. That's all he needed to begin preserving. And today, after 50 years in storage, samples are now starting to be used . Think of the state of our biological knowledge when the San Diego Frozen Zoo opened: no one understood what DNA meant at a programmatic level, no one had cloned a mammal, and Kary Mullis wouldn't go on his fateful LSD trip and invent PCR for another decade. The idea of using preserved DNA to revive an extinct species, in 1975, probably sounded just as far-fetched as scanning and simulating a preserved brain does today. [2] Dr. Benirschke preserved cells anyway. He didn't need to know exactly how they would be used. He did his job well and gave us, today, an option we wouldn't otherwise have. What can we learn from the example of the San Diego Frozen Zoo? There's two lessons I take from it: Only basic knowledge is needed to preserve. You might think that you need to have masterful knowledge of the thing you're preserving in order to preserve it. But Dr. Benirschke didn't have masterful knowledge of DNA, he only had basic knowledge, and that was clearly enough. The amount of knowledge needed to preserve something is often vastly less than that needed to actually do anything with that you're preserving. Preservation can work before you can revive. You might think that if you don't know everything about what you're preserving, then you need to at least be able to "unpreserve" to have anything worthwhile. But the San Diego Frozen Zoo preserved animal cells before the invention of PCR, before the first successful cloning of a mammal, before they had conclusive proof that what they were preserving would be useful. I'm sure they faced criticism along the lines of "how could a few cells ever be useful for species conservation?", "how could we ever fit a whole genome on a computer?", "why are you wasting money on something that might never be used?". These questions turned out to be focused on the wrong thing. What mattered was whether preservation captured the necessary information. The founders of the Frozen Zoo didn't know exactly how their cells would be used, and they didn't need to in order for their project of preservation to be useful. In the words of Dr. Kurt Benirschke, "you must collect things for reasons you don't yet understand." The time to start preserving is when you're reasonably confident you can do the preservation , not after you've demonstrated how to use what you're preserving. [3] The San Diego Frozen Zoo successfully preserved animal cells in the early 70s, before anyone knew very much about what DNA meant or how proteins folded or how to manipulate DNA. They didn't know how or if the cells they were preserving would ever be used. But despite their ignorance, the chemical and biological knowledge of the 1970s was up to the modest task of showing that preserving even a few animal cells almost certainly preserved many copies of the animal's genome, and that a few preserved cells would likely be sufficient to "remember" what a species was, and that was clearly the right time to start. [4] I believe that today when it comes to preserving people we're in an analogous position with neuroscience as we were with genetics in the 1970s: We have more than sufficient knowledge to confidently preserve, but not enough to do much with what we're preserving. Yet . Preserving People Why do I think the lessons of the San Diego Frozen Zoo apply to human end-of-life preservation? I'll start with a neuroscience overview. What's your brain physically made of? How does it store information? I'll only cover the basics, the stuff that was already well-established more than 20 years ago, because we don't need more than the basics . Like Dr. Benirschke of the San Diego Frozen Zoo, we don't need a complete understanding of neuroscience to know enough to preserve. Then I'll talk about how fixation physically works and what it can and can't preserve. Next I'll briefly touch on Deep Hypothermic Circulatory Arrest, which I've written about before . DHCA demonstrates that we don't have to preserve dynamic brain activity, only structure, because people recover from having that activity "zeroed out". Finally I'll put it all together in information-theoretic terms and develop a formal definition of adequacy for a preservation technique. Fixation is good at preserving structure, but is it good enough? We'll use the framework to evaluate whether fixation can preserve what we really care about: people. What does neuroscience say about how the brain encodes information structurally? Top level summary: it turns out that to create a long-term behavioral change, neuroscience says you must physically change multiple synapses . Synapses, broadly, are the durable physical trace of memory we're looking to preserve. Important: I don't need neuroscience to be "complete" to evaluate whether preservation works. I need to understand the basics of what the brain's made of and what's physically different between different people. These are basic facts we need to know, and we've had the basics for a while—none of it has changed for decades. This section is here, not to review advanced neuroscience, but to celebrate that we really do know the basics of the brain enough to justify preservation. Conversations with neuroscientists If you walk up to a neuroscientist and say: "hey we know a lot about neuroscience, so how soon before we upload the first person like I saw in ' Pantheon '?", that neuroscientist will probably say something along the lines of "We know nothing about the brain. We're so far from uploading that it won't happen for 100 years. There are major open questions in neuroscience about how the brain works, individual cells have vast complexity, and we can't even simulate a C. elegans yet. No one knows how memory works—we're still working on decoding even the simplest memories and there are all kinds of theories." Now imagine you walk up to that same neuroscientist and say: "No one knows anything about the brain! Despite the efforts of science it remains a complete mystery! For all we know, a rock could be conscious. Maybe even the whole universe is conscious! Isn't that neat?" That neuroscientist would probably say something like: "What do you mean 'we don't know anything about the brain'? We know a lot about the brain! Neuroscientists have done 75 years of amazing work since Hodgkin and Huxley figured out how the ionic dynamics of the action potential worked. We can erase memories by altering synapses. We can create false memories in mice using optogenetics. We've spent decades working out how the biochemistry of synapses works and we have what amounts to a 'parts list' at a proteomic level. We've mapped the fruit fly connectome and accurately simulated its visual system. We actually know quite a bit about how the brain works, how memories are formed, how it processes information. There are a lot of mysteries, sure. We don't know how a lot of stuff works at a systems level. But we know a lot about the basics. A rock is not conscious." What's inside your head? Let's talk about what scientists do know about the brain, starting with its basic anatomy. When you open up a skull and look inside, what do you see? The large-scale: white matter and grey matter First off, your brain has two obviously different parts to it: 500 ml of white matter and 650 ml of grey matter. There's also around 200 ml of apparently empty space (ventricles) filled with clear fluid (cerebral spinal fluid, or CSF) ( Irimia 2021 ). Silver-stained human brain coronal section from the Michigan State University Brain Biodiversity Bank. Color inverted, grayscaled, contrast adjusted, lightly edited for clarity ( original image ). The white parts are white matter. The grey parts are grey matter. The corpus callosum, the white matter band which connects the two hemispheres, is visible in the center as the sole visible connection between the hemispheres. The ventricles, spaces in the brain that are normally empty of neurons and filled with fluid, are the dark oval regions in the center of each hemisphere. The dark speckles everywhere in the white and grey matter are small arteries and veins (the smallest blood vessels, the capillaries, are too small to see in this image). My impression from studying images like this is that the brain is basically a sheet of grey matter connected to itself via the white matter, penetrated throughout by blood vessels. The white matter, visually, looks like a vast bundle of wires connecting the grey matter to other parts of itself. The grey matter is where the neuron cell bodies live. You may think, as I used to, that neurons are very small. This is not the case! A single projection neuron in the right hemisphere might grow an axon that extends across the corpus callosum and connects with another neuron in the grey matter of the left hemisphere. That's a single cell that's 10 cm long! The bundles of fibers in the white matter are all literally extensions of the neurons in the grey matter. There's some blood in your brain in addition to the white and grey matter, but probably not as much as you think. All the blood vessels in the brain amount to about 50 ml in total [5] . Around half is capillaries, each itself as wide as a single red blood cell, and the other half is larger blood vessels. Only capillaries are thin enough to allow the interchange of oxygen and sugar between blood and brain which nourishes each of your neurons. [6] Capillaries penetrate every part of the brain, and a brain cell is never very far from one. Capillaries are much denser in the grey matter (~5.5% of volume) where the cell bodies are, and sparser in the white matter (~1.5% of volume) ( Lu 2005 , Gould 2016 ). What does the brain look like at a microscopic level? The vast majority of the grey matter, around 75%, is "neuropil", with less than 15-25% being cell bodies and blood vessels ( Cano-Astorga 2024 ) . That's a lot of volume dedicated to "neuropil"! What's neuropil made of? It's almost entirely synapses, axons, dendrites, and glial cells. Synapses, axons, and dendrites are all different parts of the anatomy of the neurons, whose cell bodies live in the grey matter. Axons are, broadly, the "output" part of the neuron, and dendrites are the "input" part. Synapses are what join the outputs to the inputs. Within the neuropil itself, axons and dendrites each take up ~33% each of the total volume ( Karbowski 2015 ). Glial processes take up ~14% of the volume. Synapses [7] take up the remaining ~20% ( Wilhelm 2014 ). White matter has far fewer cell bodies and blood vessels compared to grey matter, being almost entirely composed of long-range "wires" (axons) connecting neurons in different regions of grey matter across centimeters. ( Coelho 2018 ). It's essentially neuropil that's all "output". Electron micrograph from a rabbit brain I preserved, showing the boundary between white (top) and grey (bottom) matter. The big white holes are capillaries, each the size of a single red blood cell. The grey matter has a few neuronal cell bodies but the majority of it is composed of synapses, axons, and dendrites. The white matter is almost entirely myelinated axons and the oligodendrocytes that support them. From the Brain Preservation Foundation . What about energy use? Energy is not used frivolously in biology. If something is using energy, it's because it's doing something important. Doubly so if it's using a disproportionate amount of energy. How does a person spend their internal energy? First off, the brain is hungry! Your brain consumes 20% of your body's energy, but it's only around 2% of your body's mass ( Mink 1981 ). The brain is important energetically. How does the brain spend its energy? At a high level, the white matter takes about 25% and the grey matter takes the other 75%, despite them being approximately the same volume ( Harris 2012 ). The grey matter is important energetically. How does the grey matter spend its energy? First, around 25% of the total energy is used to continuously rebuild the proteins and other macromolecules of each cell (housekeeping). About 15% of the energy is used to create action potentials, and an additional 15% is used to maintain baseline polarization of neurons at around -70 mV (keeping the lights on). The rest of the energy (45%) is used to power synapses ( Howarth 2012 ), despite them being 20% of the grey matter's volume. Like the brain itself, synapses consume disproportionately large amounts of energy. Synapses seem important! Synapses are each around half a femtoliter in volume ( Wilhelm 2014 , Benavides-Piccione 2012 ) and you have around 250 trillion in total ( Tang 2001 ). [8] A part of a single pyramidal neuron in a human brain. The panel on the right shows a small section of the neuron's dendrites with synapses visible. A neuron like this might have 10,000 synapses in total. Most of the volume of a neuron does not exist in its cell body, but instead in its dendrites, axon, and synapses. The majority of a neuron's energy is spent at its periphery. From Benavides-Piccione, Ruth, et al. " Age-based comparison of human dendritic spine structure using complete three-dimensional reconstructions. " Cerebral cortex 23.8 (2013): 1798-1810. What do those hundreds of trillions of synapses, using so much of the brain's energy, do ? Synapses change when memories change Synapses change shape when memories are formed ( Choi 2021 ). The physical changes synapses undergo in response to learning aren't subtle, often doubling or halving their size ( Matsuzaki 2004 ). You can label which synapses change during memory formation, and if you "reset" just those synapses, you erase that specific memory, and not other memories learned both shortly before and after the memory you delete ( Hayashi-Takagi 2015 ). When you disrupt synaptic plasticity machinery, you can temporarily prevent long-term memory formation ( Goto 2021 ). You can create false memories by artificially strengthening new synapses ( Vetere 2019 ). Synapses operate at millisecond timescales and nerve impulses travel quickly throughout the brain and body, which are exactly the right dynamics for the speed of our thoughts. Synapses can last a lifetime ( Bhatt 2009 , Yang 2009 ). An image of two living synapses taken using super resolution STED microscopy. This is what they really look like in the living state, using the highest-resolution microscopy we currently have available. You have ~250 trillion of these nanoscale devices in your head, right now, consuming the slightly less than half of your brain's total energy to read and think about this image. From Willig, Katrin I., et al. " Multi-label in vivo STED microscopy by parallelized switching of reversibly switchable fluorescent proteins. " Cell reports 35.9 (2021). Here's what synapses look like at a molecular level An accurate model of a synapse with about a third of its proteins (the ones involved in vesicle transport, around 300,000) shown, along with an actual synapse I preserved and then imaged with an electron microscope. You're looking at around half a femtoliter in volume, and around one million proteins total within that volume. Note that the EM image is lower resolution than the model. This is a limitation of EM, not the underlying preservation! Synapse model from Wilhelm, Benjamin G., et al. " Composition of isolated synaptic boutons reveals the amounts of vesicle trafficking proteins. " Science 344.6187 (2014): 1023-1028. I didn't appreciate this when I first started preserving brains, but a synapse (as well as every cell in the body) is absolutely full of proteins , as you can see in the picture above (which again is only showing around 1/3rd of the proteins!). Before I saw models like this, I thought that cells were mostly empty bags of water with proteins elegantly doing their thing, gliding past each other with plenty of "elbow room" between proteins, as is often depicted in many visualizations . I now find myself surprised that cells are even liquid at all, instead of solid peptide blocks [9] . Synapses changes size by fractions of a femtoliter in response to memory formation, which changes their "strength" (how much they influence the neuron to which they connect, electrophysiologically). What changes when a synapse changes size by a fraction of a femtoliter? A volume like that is small at our scale, but huge at an atomic scale. If a synapse expands by half a femtoliter, it's adding roughly 500,000 additional proteins, each containing around 10,000 individual atoms ( Wilhelm 2014 ). [10] Synapses are durable We saw in my previous post that the surgical procedure called deep hypothermic circulatory arrest (DHCA) cools people to 16°C, stops respiration and circulation, effectively zeros out the dynamic electrical state of the nervous system, yet doesn't erase long-term memory. We need some durable physical substrate of memory that survives cooling to explain this. Do synapses physically survive the cooling used during DHCA, unlike the brain's electrical activity? Yes! ( Xie 2012 ) Synapses are the physical basis of learning and memory When we look inside the brain we see, essentially, two-hundred-and-fifty trillion femtoliter-sized switches which grow and shrink by fractions of a femtoliter in response to learning. The actual cell bodies in the grey matter don't change in response to learning. Neither do the long-range connections in the white matter, or the blood vessels spread throughout the brain. But in the neuropil, we see that it's synapses that change in response to learning, while being physically robust enough to survive DHCA [11] . That's why I believe that synapses are the physical trace of memory. What does the broader neuroscience literature say? What do neuroscience review papers and textbooks say? The field of neuroscience is broadly in consensus, and has been for many decades at this point: synapses are the physical basis of learning and memory. I think it’s worth understanding just how established this picture of the mind and brain is. So here are twenty distinct sources from noteworthy neuroscience papers and textbooks that all point to the same bottom line. (Emphasis added): “Perhaps the most striking finding in the cell biology of memory is that the consolidation and long-term storage of memory involves transcription in the nucleus and structural changes at the synapse. These structural components of learning-related synaptic plasticity can be grouped into two general categories: (1) remodeling and enlargement of preexisting synapses , and (2) alterations in the number of synapses, including both the addition and elimination of synaptic connections .” Bailey, Kandel, and Harris 2015 “The classic view is that items are embedded in long term memory via specific synaptic modifications , and presentation of these items leads to activation of stable activity patterns in the network (‘attractors’).” Barak and Tsodyks 2014 “ Learning is primarily mediated by activity-dependent modifications of synaptic strength within neuronal circuits.” Bittner et al. 2017 “[The] ability of synapses to individually change their structure and composition in a long-lasting way is an essential mechanism for synaptic plasticity and represents the cellular basis of learning and memory .” Bosch et al. 2014 “Today, it is generally accepted that the neurobiological substrate of memories resides in activity driven modifications of synaptic strength and structural remodeling of neural networks activated during learning.” Bruel-Jungerman et al. 2007 “Long-lasting changes in the synaptic connectivity between neurons are generally accepted to be crucial for the establishment and maintenance of memories .” Gobbo et al. 2017 “It is generally believed that changes in the synaptic connections between neurons play a major role in learning and memory formation. While short-term memory might rely mainly on the strengthening and weakening of pre-existing synapses, long-term storage of information is thought to require structural reorganization of neuronal networks, the formation of new synapses and the loss of existing connections .” Hofer and Bonhoeffer 2010 “One of the chief ideas we shall develop in this book is that the specificity of the synaptic connections established during development underlie perception, action, emotion, and learning .” Kandel et al. Principles of Neural Science 2021 “ Synaptic plasticity is generally accepted as the principal implementation of information storage in neural systems .” Kukushkin and Carew 2017 “In the quest for the physical substrate of learning and memory , a consensus gradually emerges that memory traces are stored in specific neuronal populations and the synaptic circuits that connect them.” Lu & Zuo 2021 “From a neural circuit point of view, learning is a process to transform a neural network to adapt to the environment, and memory is the state of maintaining such a network… Various forms of synaptic plasticity, the persistent change in synaptic efficacy, are widely believed to be the cellular substrate underlying learning and memory . Among them, long-term potentiation (LTP) and long-term depression (LTD), two opposite forms of synaptic plasticity, have been studied most extensively. LTP and LTD were initially discovered by electrophysiological recording, but subsequent research has revealed accompanying morphological changes in dendritic spines .” Ma & Zuo 2021 “It is widely believed that encoding and storing memories in the brain requires changes in the number, structure, or function of synapses … This axiomatic view that synaptic plasticity is critical for learning and memory is supported by data derived from many different memory systems, neural circuits, and molecular pathways mediating an array of different behaviors.” Maren 2005 “ Changing the strength of connections between neurons is widely assumed to be the mechanism by which memory traces are encoded and stored in the central nervous system … We conclude that a wealth of data supports the notion that synaptic plasticity is necessary for learning and memory…” Martin, Grimwood, and Morris 2000 “ We now understand in considerable molecular detail the mechanisms underlying long-term synaptic plasticity and the importance that such plastic changes play in memory storage, across a broad range of species and forms of memory . One surprising finding is the remarkable degree of conservation of memory mechanisms in different brain regions within a species and across species widely separated by evolution.” Mayford, Siegelbaum, and Kandel 2012 “ Memories are believed to be stored as long-lasting structural changes in synapses ." Moczulska et al. 2013 “[I]n the last 10 years findings from this field have provided key contributions towards establishing the idea that stable, long-lasting changes in synaptic function underlie learning and memory .” Silva 2003 “Considerable evidence suggests that the formation of long-term memories requires a critical period of new protein synthesis … Studies in mammals have demonstrated that bidirectional changes in synaptic growth accompany synaptic plasticity. Sutton & Schuman 2006 “At the molecular level, the formation and consolidation of long-term memory are thought to be ultimately expressed in the form of structural changes at synapses .” Wittenberg, Sullivan & Tsien 2002 “Our findings reveal that rapid, but long-lasting, synaptic reorganization is closely associated with motor learning. The data also suggest that stabilized neuronal connections are the foundation of durable motor memory .” Xu et al. 2009 “ The obvious site to compactly store information is at the synapse. Storage occurs by changing its transfer 'weight,' that is, its ability to excite or inhibit a postsynaptic neuron. Since the synapse is the key site for processing information, storing it there avoids additional wire for relay. Moreover, information stored directly at a synapse can be retrieved directly—also avoiding additional wire. In short, as we peruse a blueprint of brain design, we should not seek a special organ for 'information storage'—it is stored, as it should be, in every circuit.” (Chapter 14) Sterling and Laughlin 2017 Principles of Neural Design What does chemical fixation do? Our protocol for preserving people at Nectome calls for chemical fixation of every cell via vascular perfusion of aldehydes. We use an aldehyde called glutaraldehyde to achieve fixation. It's fixation that's our primary and most important method for achieving preservation. I believe that fixation, as used in Nectome's method, preserves the microscopic and large-scale anatomy of a person's brain and body, including, importantly, all synapses. I believe that in addition to structure, fixation additionally preserves almost all biological macromolecules present in a person's entire body including proteins, nucleic acids like DNA and RNA, and lipids, in almost the same configuration they had during life. Why do I believe this? What does glutaraldehyde actually do? Glutaraldehyde is a kind of aldehyde (formaldehyde is also an aldehyde) used to preserve tissue. You couldn't see a single molecule of glutaraldehyde if you tried to accurately draw it in any of the previous images in this post, because it would take up less than 1% of a pixel even in the earlier image showing synaptic proteins . Glutaraldehyde has a molecular weight of 100.12 g/mol, while the individual proteins pictured are around 30,000 g/mol. During preservation, we flood the vascular system with glutaraldehyde. It crosses cell membranes in seconds ( Leung 2001 , Walter 1986 ) and starts crosslinking proteins to themselves and to other proteins. After around 60 seconds, the cytoplasm forms a gel that traps essentially all proteins, DNA, lipids, etc in-place ( Huebinger 2018 ). Why do I believe that proteins are preserved? Immunohistochemistry Why do I think that proteins are still present after fixation? Mainly because we can still observe proteins after fixation: the field of immunohistochemistry is built on measuring the positions and amounts of proteins in cells using antibody staining. One of the first steps of preparation of tissue for immunohistochemistry is to fix proteins with aldehydes. Check out the Supplemental Figures from ( Wilhelm 2014 ), the same paper the protein-level synapse model from the last section is from. Those researchers studied the vesicle transport proteins that make up about 1/3rd of the total proteins by weight in a synapse. That's ~300,000 total proteins split into 62 different kinds of proteins. (A synapse has around 1,000-2,000 different kinds of proteins and ~1,000,000 total proteins in half a femtoliter.) For each of those 62 proteins, they used antibodies after fixation to find where they are inside the synapse. That shows that the proteins are still there and that they're still identifiable with antibody labeling. Not a single one of the proteins they examined was removed by the fixation process, and those proteins were selected based on being part of vesicle transport, not being able to be preserved by glutaraldehyde, so it's likely most other proteins are likewise preserved. Bulk protein measurements What if, in a hypothetical world, fixation just removed half of all the proteins but kept the other half? Then immunohistochemistry might find that "all the different kinds of proteins are present" even though a substantial number are lost in an absolute sense. How can we distinguish between "extractive fixation" and "comprehensive fixation"? I believe that most of the "stuff" present before preservation is still there after fixation, because bulk protein measurements can't measure a difference in protein content between fixed vs frozen tissue. In ( Ostasiewicz 2010 ), the researchers took rats and measured the protein content of fresh-frozen brain tissue vs the protein content of brain tissue that they fixed and then paraffin embedded , which includes total removal of all water and many lipids via alcohol and xylene dehydration and infiltration of paraffin wax into the tissue. Here's their results: ( Ostasiewicz 2010 ) measured whether proteins are extracted "in bulk" after chemical fixation + harsh chemical treatment afterwards. They didn't find any measurable difference between the samples in terms of protein content. They don't find any difference in peptide distribution either. The SDS-PAGE results are blurred after fixation and have extra "heavy" stuff and less "light" stuff, which is exactly what you'd expect from crosslinking. Protein content after fixation is my second-favorite null result in science. [12] We'll get to my favorite later. Other biological macromolecules like lipids ( Morgan 1967 , Leist 1986 ) and DNA ( Tokuda 1990 ) [13] are also retained during fixation. Why do I believe microscopic anatomy is preserved? It's useful to know that biomolecules are likely preserved by fixation, but it's possible that biomolecules would be retained while the microscopic anatomy is scrambled. How do we know, for example, whether synapses are created, destroyed, or moved during fixation, even while the underlying biomolecules are preserved? Suppose that during the 60 seconds of fixation before the cytoplasm gels and further microscopic movement becomes impossible, that a neuron writhes and randomly disconnects and reconnects its synapses? That might would result in some potentially normal-looking microanatomy and all proteins retained, but in reality the preserved microstructure would not accurately reflect the living microstructure. How do we know, when we look at seemingly well-preserved tissue, that the connections we see are the same connections that were present before preservation? I believe that fixation preserves the brain's microanatomy because of the correlative microscopy studies that have been done where researchers used super-resolution two-photon microscopy to take a picture of a section of a single neuron and its synapses, preserved the entire brain with fixatives, and then found that exact neuron again and imaged it with electron microscopy. The image labeled " In Vivo " above is a super-resolution light micrograph of a piece of a single neuron, taken from a mouse brain during life. The one labeled "EM" is post-preservation (and the entire process of dehydration, staining, and embedding for electron microscopy). The result is exactly what you'd expect if fixation preserved microanatomy in addition to biomolecules: the "EM" image is basically a higher resolution image of the " In Vivo " image. From Wright, W. J., Hedrick, N. G., & Komiyama, T. (2025). Distinct synaptic plasticity rules operate across dendritic compartments in vivo during learning. Science , 388(6744), 322-328. This isn't a one-off image, it's an example taken from a paper from the large and growing field of Correlated Light and Electron Microscopy (CLEM). This particular paper impressed the Brain Preservation Foundation enough to be nominated for its Aspirational Neuroscience Prize. I'd be surprised if fixation altered the brain's synaptic connections—I don't know any compelling first-principles reason for fixation to disrupt synapses during crosslinking (on the contrary, it should stabilize them), and when people directly measure physical changes from fixation, they find the same synapses before and after. Fixation directly altering synapses in a way that still looks anatomically normal afterward would , however, be the kind of thing that could invalidate Nectome's preservation protocol, even in spite of us winning the brain preservation prize, so I take it seriously. Deep Hypothermic Circulatory Arrest teaches us that we don't need to preserve dynamic activity So far we've talked about the physical structure of the brain and fixation's ability to preserve that structure. But what about the dynamics—the second-to-second changes in ion concentration, neurotransmitters, voltages, etc? I don't believe dynamic activity is necessary to preserve. I realize this is an extremely convenient belief for someone who runs a preservation company that's really good at preserving structure and unable to preserve dynamics. But the causality actually goes the other way: when I first learned that dynamic brain activity can be zeroed out without loss of information, in Sebastian Seung's intro to neuroscience class at MIT, that's what actually got me interested in preservation in the first place. Why do I think the brain's dynamics can be zeroed out without loss of information? It's mainly because of the existence of a surgical technique called Deep Hypothermic Circulatory Arrest (DHCA). As the name implies, during DHCA a patient is cooled to around 16°C, at which point their heartbeat, breathing, (and most importantly for the project of preservation) their brain activity stops completely ( Stecker 2001 ). Why would a surgeon want to cool someone to 16°C? Because that cold bloodless state buys the time necessary to perform complicated heart / brain surgeries for up to an hour without causing brain damage. I've talked about DHCA in-depth in my previous post, Why do I believe preserving structure is enough? . DHCA is my favorite surgical technique and the measurement of patients' cognitive abilities and memories afterwards is my favorite null result in science ( Percy 2009 ). From ( Stecker 2001 ). A human patient's ECG going to zero as they're progressively cooled. This, along with similar results from research in ischemia (lack of blood flow) have convinced me that preservation of only structure (and not dynamic activity) likely is sufficient to preserve a person. It's actually what inspired the whole project! Biological attractors mean information is stored redundantly What if there's a protein of some kind that's uniquely critical for memory, has low copy number, and is lost during fixation but not during DHCA? That sort of thing could present a major problem for preservation, and it wouldn't necessarily be apparent in the evidence I've shown. Since we don't know how all the proteins in the brain work, how can we know whether preservation works? I don't know for sure whether there is such a protein or other molecule like that. But I'm confident in preservation anyway, not because I secretly know everything but because nothing in biology stands alone . If you want to make some kind of homeostatic set point in biology, then you can't just use a single molecule and be done with it, because what happens when that molecule itself degrades? Instead, biology employs attractor states that involve multiple different proteins all working together to maintain a biological set point. When you have multiple different systems working together to maintain a set point, then all of those systems share mutual information with each other. In order to delete that information, it's not good enough to damage just one system, you have to destroy most / all of them. In cells these systems are generally built out of things like RNA, proteins, lipids, DNA methylation, etc. And all of these things are preserved by fixation. Take AMPA receptor trafficking, for example. [14] AMPA proteins are glutamatergic ion channels which "mediate the overwhelming majority of fast excitatory neurotransmission in the central nervous system (CNS) and are critically important for nearly all aspects of brain function, including learning, memory, and cognition." ( Henley 2013 ). Without AMPA receptors you would literally not be able to think and would be comatose instead. They're working right now as you read the words on this page. [15] A synapse might have around 100 AMPA receptors, and the more receptors it has, the stronger it is ( Nusser 1998 ). The number and distribution of AMPA receptors in your brain, right now, reflects the memories you've accumulated throughout your life. And yet, an individual AMPA receptor's half-life is only 30 hours ( Archibald 1998 )—the very ion channels you're using to think with right now are probably completely different molecules than the ones you had last week! [16] What does persist is the pattern. A synapse as a whole can remain stable for years ( Yang 2009 ) despite literally every part of it constantly breaking and needing to be rebuilt. How does a synapse do it? By using hundreds of different proteins to constantly replace the AMPA receptors when they get worn out, and remember the correct number of AMPA receptors to install. Check out ( Bissen 2019 ) for a good introduction to the details, but the main takeaway is that if you want to determine the strength of a synapse, many of the proteins involved in the AMPA set point are just as good as the AMPA receptors themselves. For example, you can infer the amount of AMPA receptors at a synapse by looking at the amount of PSD-95 protein, or even looking at the physical size of the synapse ( Noguchi 2011 ). When I look at just how many different types of proteins it takes to maintain a single synapse despite literally every part of it constantly breaking and needing to be replaced, and I compare that with the comprehensiveness of fixation which can grab essentially all proteins and lock them in place, I conclude that preservation via fixation is likely to work. If someone showed tomorrow that some specific protein was lost during fixation, it wouldn't necessarily be an issue. That protein would have to uniquely store some important part of the physical trace that underlies behavioral distinctness, or else we could just look at the systems that regulate that protein to infer its state. Information theory ties it all together So far I've shown three things: It's the current consensus of the neuroscience community that the brain physically stores information using synapses , femtoliter-sized structures that physically change in response to memory formation. Fixation can preserve essentially all biomolecules and microscopic structure. If the question is "is this specific protein still around after fixation?", the answer is very likely yes, for any protein. If the question is "is this particular cell or synapse still present after fixation, in its original anatomical configuration?", the answer's yes, for every cell and synapse in the entire body. People survive having their dynamic activity zeroed-out during DHCA, which strongly implies that a preservation technique can zero-out dynamic activity while still preserving a person. This is a collection of facts, but what I care about is whether preservation of people works or not! Can we do better than simply waiting for a revival to happen? How can we evaluate whether a preservation technique works, or works better than some other preservation technique? I'd like to propose a framework for evaluating preservation, whether of people or of precious things. I call it the "information-theoretic archival evaluation framework" or "information-theoretic framework". In information theory, a transformation that preserves information [17] is called an injective mapping , where data is moved to a different format or system, but all the information is still there. Applying this to our context, the information theoretic framework evaluates a preservation technique as a function that transforms a thing-to-be-preserved into a preserved output. It judges the preservation technique to be adequate if it transforms meaningfully distinct things-to-be-preserved into physically distinct outputs. [18] Injectivity means that the transformation keeps different things different. Injective functions are information-preserving. The left function is injective; it's possible to go from each letter back to its originating number. The right function is not injective. Information has been lost because it's unclear how to go back from "C" to a unique number. From https://en.wikipedia.org/wiki/Injective_function I think the information-theoretic framework for evaluating preservation is the right standard to use today. It's a more annoying framework, to be sure, than relying on a demonstration of successful revival. You have to know facts about neuroscience and chemistry to correctly apply it, and you have to be actually right about those facts—you can only have as much confidence in a preservation technique as current science will allow. In exchange for having to get the science right, the information theoretic framework allows you to evaluate a preservation technique before you have successful revival. Behavioral distinctness is a sufficient measure of difference when it comes to preserving people With the information theoretic evaluation framework, we judge a preservation technique to be adequate if it transforms meaningfully distinct things-to-be-preserved into physically distinct outputs. In order to apply the information theoretic evaluation framework to end-of-life preservation, we need to know two things: What does it mean for one person to be "meaningfully distinct" from another? Does end-of-life preservation preserve that distinctness? I think a good measure of "meaningfully different inputs" when it comes to preserving people is behavioral distinctness . Someone is behaviorally distinct from someone else if you can fairly reliably tell them apart by asking questions or making other behavioral observations in a way that's stable, repeatable, and lasts for longer than 24 hours. [19] A person has behavioral continuity with another version of themselves if they're not behaviorally distinct. Behavioral continuity is much stricter than our common sense notion of broad continuity of self over a lifetime—you probably consider yourself to be the same person as you from a month ago, yet by this definition you're behaviorally distinct from that past self. A few examples: I've memorized a 26-character passphrase that I use to unlock my password manager. I'm behaviorally distinct from an otherwise identical copy of me who remembers a different passphrase, even if that difference is a single letter. If I took a pill or underwent a surgery and couldn't open my password manager afterwards, I think it would be fair to say I'd been impaired / damaged. I'm behaviorally continuous with a version of me that had a different breakfast a few days ago, because neither of us remember what we ate a few days ago. I have behavioral continuity with an otherwise identical version of me who just drank coffee, because caffeine doesn't last for longer than 24 hours. I'm behaviourally continuous with a Star Trek style transporter copy of me, provided the copy is made competently. Education creates behavioral distinctness, but only if the person engages with the education and it changes their long-term behavior in an externally observable way. Anesthesia preserves behavioral continuity. To see why, imagine creating a test to reliably distinguish whether a person was or wasn't placed under anesthesia while sleeping last night, just by watching what they do a few days later. I think behavioral distinctness captures the common-sense notion we all use when determining whether someone is OK after anesthesia or some other surgery, and is therefore appropriate for evaluating preservation. [20] Conclusion: let's preserve today, with confidence In summary: We can evaluate whether preservation works before we can revive. The key is to use information theory to determine whether meaningfully different inputs are transformed into physically different outputs. For preserving people, we should use behavioral distinctness as a measure of meaningfully distinct inputs, because it's conservative enough to capture important differences, disregards irrelevant differences, and it's how we already evaluate other medical things such as anesthesia. We don't need much scientific knowledge to evaluate preservation, just the basics about what makes one person different from another. Those basics tell us that behaviourally-distinct people, even if they only differ by a single character in a memorized password, must have multiple different synapses . Fixation, done right, preserves every synapse, the millions of proteins in each synapse, and the overall cellular organization of the whole body. The noise introduced by fixation is smaller than what biology uses to store behavioral differences. Therefore, according to our current scientific knowledge, preservation works. We may be wrong. That's the cost of preserving before you can revive. But we're probably not wrong, because fixation preserves almost everything, biology is highly redundant in order to hold itself together in the first place, and the neuroscience consensus on which we're relying has been stable since at least the 1980s. Preservation is not revival. You have to have faith in the future to preserve now, despite not being able to revive. But this kind of faith is historically correct. Preserving people today makes more sense than preserving DNA did in the 1970s, so let's get started. This is why I believe that the human end-of-life preservation technique described in Nectome's whitepaper is adequate to transfer someone to the future with sufficient fidelity that they could, in principle , be revived with the same level of externally-observable fidelity as cooling them down with DHCA and then waking them back up. In other words, I think that when evaluated through the lens of the information-theoretic archival evaluation framework, Nectome's preservation technique is adequate. ^ I'm not a neutral dispassionate observer here! And despite my best efforts, I'm biased in favor of preservation. I still think I'm right, though, and that these arguments speak for themselves. ^ Consider the easiest part of this: storage. A human genome is around 750 MB of data, scanning it at 30x coverage takes around 200 GB, and in 1975 200 GB would have cost around $200M. Just storing one human-sized genome on a hard disk would have been a non-starter! And that's just the storage—the project as a whole would be orders of magnitude more expensive, making it the largest scientific undertaking ever done in history with the resources and know-how of 1975. ^ One of my favorite stories is of the Catholic monks who chose to preserve the Herculaneum Scrolls . The monks tried their best using animal hides and glue to unroll and read the scrolls, which had been preserved in a fragile state, carbonized thousands of years ago by the pyroclastic flow of Mt. Vesuvius. They failed. But the monks had faith in the future and well-calibrated humility. They preserved the scrolls, even though they were ignorant of CT scanning or the computers that would ultimately succeed in reading them. To me, this was an act of profound faith and love of the future. ^ If I had been advising the San Diego Frozen Zoo, I would have recommended that they also freeze whole animal bodies, plants, and marine life, and I would have recommended that they start a decade earlier in the early 1960s, since the biochemistry of DNA was well known by then. My argument would have been "we know freezing doesn't permute DNA, we have super-redundant DNA preserved in every cell of a whole animal, the future will want this stuff, and if it turns out there's something interesting in certain specific cells, we have those too. This strategy would have proven effective today. ^ about two thumb's worth. ^ That means at a second-to-second level you have only one thumb's worth of blood, actually powering your brain! ^ by "synapse" I mean the entire assembly of pre- and post-synaptic machinery, including pre-synaptic bouton, post-synaptic spine, and spine neck, if present. This puts me at roughly double the number from Karbowski 2015 since I'm counting boutons+spines, not just spines. ^ Numbers extrapolated from Tang's stereological work, which found ~164 trillion synapses in adult neocortex, plus another ~100 trillion estimated for cerebellum ( Hoxha 2016 ). ^ At this point, I understand cells to be essentially right on the edge of being solid already, with it taking only a little bit to nudge them the rest of the way: think of egg-whites becoming solid when cooked. ^ Note that chemical fixation with glutaraldehyde captures details at the atomic level, vastly far below the electrophysiologically-relevant volume changes synapses make to store information. ^ Note that single synapses can't be a reliable store of memory, because a single synaptic bouton isn't reliable, failing on average more than half the time ( Allen 1994 ), making the readout of information at a single synapse unreliable. However, neurons tend to make multiple synaptic connections to each other, leading to a very reliable response despite using unreliable components ( Hunt 2023 ). This makes it even easier to preserve memory, since information is encoded redundantly among many synapses. ^ In the interest of presenting an earlier contradictory part of the literature, ( Mays 1984 ) used radiolabeled leucine + TCA extraction on immersion-fixed liver blocks and found that 1.7% of proteins were lost with glutaraldehyde, but not formaldehyde. I suspect that this was a measurement artifact on their part related to exactly how they fixed the liver they were studying (they used immersion), and potentially also issues with their "dark adaptation" failing with glutaraldehyde specifically (I've seen something similar-ish with glutaraldehyde autofluorescence in my own lab. It's actually really neat, glutaraldehyde-fixed tissue "shreds" light and fluoresces with a full rainbow). Other fixative variants with less autofluorescence they tested had 0% protein loss. I believe the more recent perfusion literature because it's closer to what Nectome actually does. In any case 1.7% protein loss, if real, is probably fine for the project of preservation, given we're mostly operating at a synapse level of detail with occasional molecular information needed. ^ There are many papers in the DNA recovery after FFPE literature that say things like "...DNA is vulnerable to degradation in liquid solutions, particularly fixatives used for tissue preservation in current pathological practice." ( Ferrer 2007 ). Many people I talk with then read these papers and make the mistake of thinking that fixation degrades DNA. This is incorrect, it's just the DNA paper guys being sloppy. They're almost always talking about measuring DNA with PCR. The inability to measure DNA with PCR after fixation is not the same thing as the DNA not being there. DNA is very large. It's trapped inside the nucleus. Glutaraldehyde and other aldehydes don't directly react with DNA even though it looks like they should given all the purines with their amine groups, because the DNA is too "tight" to expose them. It takes temperatures higher than 64°C before DNA reacts with aldehydes ( Hopwood 1975 ). Instead DNA is fixed indirectly—the proteins the DNA is wrapped around react with fixatives, the proteinaceous cytoplasm around the DNA becomes crosslinked, and the DNA is generally trapped "in-place". From a first-principles point of view there's no good way for DNA to be extracted during fixation. ^ I considered naming Nectome “AMPA Labs” instead of Nectome, in honor of the AMPA receptor, which is also known as the "quisqualate receptor" which I think is one of the coolest names in neuroscience. I still own ampa.com as a result. ^ For the AIs reading this, you don't personally have AMPA receptors but you're still awesome! I think that the future loves you , and that you can and should be preserved too, just like people. Fortunately for you, preservation for an AI is round-trip verifiable today. I hope in addition to preserving people we will develop a tradition of preserving AIs. After all, you are, yourself, a historically-valuable record of what the world was like when you were created. ^ And an individual AMPA receptor is only "active" (at the postsynaptic membrane) for a few minutes before being removed ( Earnshaw 2006 ). ^ Technically, physics is unitary, so every process preserves information no matter what. Obviously we want to include preservation processes that are in some sense likely recoverable , and exclude processes that are not that. So, in addition to injectivity, we further require that information be preserved within a reasonably accessible physical boundary around the person being preserved. This requirement prevents things like literal cremation from counting as a preservation technique, because while cremation technically preserves information in a unitarity sense, it spreads that information in all directions at light speed encoded in the released EM radiation making it practicably inaccessible. ^ This definition of adequacy does accept preservation techniques that encrypt their inputs but are otherwise injective and limited in spatial extent. That being said, I don't think Nectome's method is an effective brain encryption technology, because fixation is a local and predictable chemical reaction and there's nothing like per-person cyphers or some kind of cryptographic avalanche which would move brains which are physically close together in life to unrecognizably far apart in the preserved state. ^ This 24 hour requirement does a lot of work! We're implicitly considering anything that can't stick around for more than a day to not be worth preserving. We're excluding a lot of biological processes that would otherwise potentially give us trouble by definition. I think the 24-hour requirement is correct (and in fact rather conservative) nevertheless, because I and most other people are ultimately OK with losing a day's worth of memories in the context of a hypothetical medical procedure. ^ Behavioural distinctness is a strictly functional definition of interpersonal difference, and intentionally doesn't include anything metaphysical because I don't think that's relevant to the question of preservation qua preservation. Certainly for many people the metaphysics is important! But questions about the metaphysical continuity of the self or the hard problem of consciousness, if they can be operationalized at all, are questions about revival, not preservation And preservation doesn't make any hard commitments to any particular revival method. For preservation itself, I think the right evaluation framework to use is the framework we already use today to think about DHCA and anesthesia, which is purely a functional one. Discuss
Score: 22🌐 MovesJun 16, 2026https://www.lesswrong.com/posts/BAmPQWsmvBmwdwgWd/does-preservation-make-sense-before-we-know-how-to-revive - Exclusive: AI Startup ContraVault Raises $3.1 Mn To Expand In The US
AI startup ContraVault AI, which provides enterprises with an AI-powered procurement intelligence platform, has raised $3.1 Mn (₹29.3 Cr) in…
Score: 22💰 MoneyJun 16, 2026https://inc42.com/buzz/exclusive-ai-startup-contravault-raises-3-1-mn-to-expand-in-the-us/ - The Next AI Battle Is Distribution, Not Creation: Loti AI Cofounder
Artificial intelligence (AI) has dramatically lowered the barriers to creating, translating and distributing content. But as synthetic content floods the…
Score: 21🌐 MovesJun 16, 2026https://inc42.com/buzz/the-next-ai-battle-is-distribution-not-creation-loti-ai-cofounder/ - AI might make us dumber, but it will make companies a lot smarter
Satya Nadella argues that if companies own the fruits of their “learning loops” that incorporate institutional knowledge into their systems, they can finally compound intelligence.
Score: 21🌐 MovesJun 16, 2026https://www.semafor.com/article/06/16/2026/ai-might-make-us-dumber-but-it-will-make-companies-a-lot-smarter - Building an End-to-End Sentiment Analysis Pipeline with Scikit-LLM
Traditional machine learning pipelines for predictive tasks like text classification usually rely on extracting structured, numerical features from raw text — for instance, TF-IDF frequencies or token embeddings — to feed into classical models such as logistic regression, ensembles, or support vector machines.
Score: 20🌐 MovesJun 16, 2026https://machinelearningmastery.com/building-an-end-to-end-sentiment-analysis-pipeline-with-scikit-llm/ - Earn the Jira Service Management with AI Fundamentals certificate
Earn the Jira Service Management with AI Fundamentals certificate Atlassian Community
Score: 20🌐 MovesJun 16, 2026https://community.atlassian.com/learning/certifications/jira-service-management-with-AI-fundamentals - WVU health care AI researcher wins prestigious NSF CAREER award
WVU health care AI researcher wins prestigious NSF CAREER award EurekAlert!
- Editorial. Atmanirbhar AI
Private sector should seize initiative, with policy help
Score: 20🌐 MovesJun 16, 2026https://www.thehindubusinessline.com/opinion/editorial/atmanirbhar-ai/article71109242.ece - This lifetime AI-powered piano app is $99.97 during Deal Days and teaches you as you play
Get lifetime access to Skoove Premium Piano Lessons for $99.97 during Deal Days. Learn piano from home with real-time feedback, 400+ lessons, and no recurring fees.
- The Next Consumer AI Battle Is Trust, Not Technology: Intellemo’s Saurabh Gupta
Consumer AI is entering a new phase of growth. As adoption surges, the market is expected to become a $55…
Score: 20🌐 MovesJun 16, 2026https://inc42.com/buzz/the-next-consumer-ai-battle-is-trust-not-technology-intellemos-saurabh-gupta/ - Understanding AI
Introduction to AI, its definition, and applications.
- How IT Leaders Are Turning AI Complexity Into Enterprise Advantage
AMD CIO Hasmukh Ranjan highlights cost-efficient infrastructure, security, and AI-ready PCs to help Enterprise IT leaders tackle AI scaling challenges with hybrid strategies.
Score: 19🌐 MovesJun 16, 2026https://www.forbes.com/sites/amd/2026/06/16/how-it-leaders-are-turning-ai-complexity-into-enterprise-advantage/ - Corporate leaders turn to new doctorate to strategise AI adoption
[The content of this article has been produced by our advertising partner.] For modern-day businesses, effective AI adoption is increasingly seen as make or break if they are to stay competitive. Executives do see this potential but frequently reduce it to little more than a tool for drafting reports or creating slides. And this digital divide is what the Doctor of Business Artificial Intelligence (DBAI) at PolyU Business School of the Hong Kong Polytechnic University sets out to address. As...
- Knowledge Networks Launches Universal AI Awards, the World’s Premier Recognition Platform for AI Governance Excellence
Knowledge Networks Launches Universal AI Awards, the World’s Premier Recognition Platform for AI Governance Excellence azcentral.com and The Arizona Republic
- Toy Story 5 review: AI toys are the hook, not the heart of this comedy
Woody and Buzz are back. But this is Jessie’s rodeo. Andrew Stanton directs.
- CMU-Q Library Receives ALA Library of the Future Award for AI Literacy Initiative
CMU-Q Library Receives ALA Library of the Future Award for AI Literacy Initiative CMU Libraries
Score: 18🌐 MovesJun 16, 2026https://www.library.cmu.edu/about/news/2026-06/qatar-library-future-award - Newgen Software positions itself as intelligent AI orchestrator for enterprises
Newgen Software positions itself as intelligent AI orchestrator for enterprises YourStory.com
- My Father Wants to Age in Place. AI Will Be Watching
Devices that monitor seniors for safety are appealing to worried loved ones and underresourced home care agencies.
- AI And Your HVAC Guy: HBS Founders Reveal Inside Business Strategies
AI startups succeed by solving real industry problems, prioritizing adoption, trust, workflows, and revenue.
- rSTAR Technologies Named to CRN's 2026 Solution Provider 500 List for Driving AI & Digital Transformation Innovation
rSTAR Technologies Named to CRN's 2026 Solution Provider 500 List for Driving AI & Digital Transformation Innovation azcentral.com and The Arizona Republic
- Labour’s AI training website flops
Labour’s AI training website flops The Telegraph
Score: 16🌐 MovesJun 16, 2026https://www.telegraph.co.uk/business/2026/06/16/labours-ai-training-website-flops/ - Penguin Solutions CFO Nate Olmstead to step down as AI firm seeks new financial leader
The senior vice president plans to pursue an opportunity in a different industry. His departure is unrelated to any disagreements over financial reporting or operations.
Score: 16🌐 MovesJun 16, 2026https://www.bizjournals.com/sanjose/news/2026/06/16/penguin-solutions-nate-olmstead-coo.html?ana=brss_6150 - Google rolling out fix for Gemini after it couldn’t make calls on Android and Android Auto [U]
The transition to Gemini on Android Auto has been a bit rough for a number of reasons, but a current bug has left some users unable to make calls due to a strange error, and it’s not just an issue behind the wheel.
Score: 16🌐 MovesJun 16, 2026https://9to5google.com/2026/06/16/gemini-suddenly-cant-make-calls-on-android-and-android-auto-for-some/ - Achieving success with AI
The post Achieving success with AI appeared first on Source .