AI News Archive: July 15, 2026 — Part 6
Sourced from 500+ daily AI sources, scored by relevance.
- Tencent Cloud expands AI agent suite in Indonesia
Tencent Cloud said more than 150 Indonesian enterprise leaders, technology executives, and partners attended the event.
Score: 50🌐 MovesJul 15, 2026https://www.techinasia.com/tencent-cloud-taps-hong-kong-as-global-expansion-hub - AI slop is taking over the housing market – and it is causing a nightmare for home-hunters
As more and more estate agents use ‘staging’ tools to give their drab properties a glow-up, Helen Coffey argues that seeing is no longer believing when looking at homes online, and wonders if the tech is having the opposite effect on would-be buyers
Score: 50🌐 MovesJul 15, 2026https://www.independent.co.uk/life-style/ai-estate-agents-house-buying-listings-b3012758.html - SK chief says next five years will determine AI industry’s future
SK chief says next five years will determine AI industry’s future 매일경제
- The execution blueprint: Driving high-yielding outcomes in enterprise AI architecture
By Vikash Sharma, CEO SparxIT Artificial Intelligence has ceased to be just a prototype stage. Across the board, corporations have plowed funds into AI pilots, generative AI tools, predictive analytics, […] The post The execution blueprint: Driving high-yielding outcomes in enterprise AI architecture appeared first on Express Computer .
- How Nidec Is Rethinking Gear Design for Humanoid and Mobile Robots
Nidec tackles the brutal gearbox tradeoffs behind humanoid robots, from zero backlash to lighter integrated actuators. The post How Nidec Is Rethinking Gear Design for Humanoid and Mobile Robots appeared first on EE Times .
Score: 50🌐 MovesJul 15, 2026https://www.eetimes.com/how-nidec-is-rethinking-gear-design-for-humanoid-and-mobile-robots/ - KB Financial launches AI Lab for workplace AI
KB Financial Group has launched a hands-on AI talent program as part of efforts to accelerate artificial intelligence adoption across the group and integrate the technology more deeply into everyday work. The financial group said Wednesday it officially launched the KB AI Lab during its AI and data innovation seminar held Tuesday at KB Kookmin Bank's headquarters in Yeouido, western Seoul. The project-based program is designed for graduates of the highest level of KB AI Academy's AI expert nurtu
- At last, a good reason to buy an AI PC: Reining in runaway token bills
Gartner thinks we’re headed to a hybrid AI model where you offload stuff to the desktop whenever possible
- OpenAI Fails To Trademark Name In EU
OpenAI Fails To Trademark Name In EU Barron's
Score: 50🌐 MovesJul 15, 2026https://www.barrons.com/news/openai-fails-to-trademark-name-in-eu-46c41d65 - Why Vision-Language Models Are Shortsighted
Why Vision-Language Models Are Shortsighted emilie.germain… Wed, 07/15/2026 - 11:08
Score: 50🌐 MovesJul 15, 2026https://mila.quebec/en/article/why-vision-language-models-are-shortsighted - ‘Botsitting’ Is the New AI Workplace Trend That’s Frustrating 87 Percent of Digital Workers
Data reveals workers are spending hours fixing confident-but-wrong AI mistakes, leading to a much more dangerous corporate habit.
- Google’s Head of Search Shares 3 Rules Every Business Needs to Win in the AI Era
Google VP of Search Liz Reid tells Inc. what Google’s latest Search changes mean for entrepreneurs.
- The small team in Montreal trying to save the world from AI
The post The small team in Montreal trying to save the world from AI appeared first on The Logic .
Score: 50🌐 MovesJul 15, 2026https://thelogic.co/news/the-big-read/the-small-team-in-montreal-trying-to-save-the-world-from-ai/ - AI’s trillion-dollar lease overhang is off the books, not off the hook
The largest AI companies are building the most expensive infrastructure programme in corporate history. They are also moving a growing share of the financing away from their own balance sheets. That does not mean the risk has disappeared. It means the risk has moved. The new AI financing stack is built from joint ventures, special-purpose […] The post AI’s trillion-dollar lease overhang is off the books, not off the hook appeared first on e27 .
Score: 50🌐 MovesJul 15, 2026https://e27.co/ais-trillion-dollar-lease-overhang-is-off-the-books-not-off-the-hook-20260713/ - Leading the Network 2030: WBBA Officially Launches AI-Net, a Globally Authoritative Data Communications Certification
Leading the Network 2030: WBBA Officially Launches AI-Net, a Globally Authoritative Data Communications Certification The Straits Times
- This AI tool doesn't just speak languages—it invents them
Artificial intelligence isn't just capable of translating between existing languages—it can also create entirely new ones.
- Thinking Machines open sources first multimodal language model, Inkling, focused on low cost and 'resistance to censorship'
Enterprises looking to move more of their agentic AI workloads to open weights models they can customize, control and run on-premises or in virtual private clouds have a strong new contender to consider. Today, Thinking Machines—the highly capitalized American AI startup founded by former OpenAI CTO Mira Murati— released Inkling , its first major language model under an enterprise-friendly Apache 2.0 open source license , and it boasts high, if sub state-of-the-art, performance for open weights models on third-party benchmarks, specifically software engineering (77.6% on SWE-bench Verified, where it beats fellow U.S. open rival Nvidia Nemotron 3's 71.9%) and voice understanding (91.4% on VoiceBench compared to 94.4% for Gemini 3.1 Pro on high reasoning effort). Another differentiator: Thinking Machines notes that Inkling was designed "to answer directly on topics that may be subject to censorship," offering enterprises concerned about factual outputs, irrespective of controversy or sensitivity, a more trustworthy option. Coming in at 975 billion total parameters, Inkling is a natively multimodal, open-weights Mixture-of-Experts (MoE) system capable of reasoning across text, images, and audio. The weights are already available on Hugging Face and the company's own model training application programming interface (API), Tinker . Designed to balance cost against performance through a novel "controllable thinking effort" mechanism, the model represents a significant departure from the black-box scaling strategies of frontier competitors. Alongside the flagship model, Thinking Machines also announced a preview of Inkling-Small, a lighter 276-billion-parameter alternative optimized for workloads where low latency and cost are paramount. Benchmarks Show a Powerful, High-End, Sub State-of-the-Art Model While Inkling is a formidable multimodal engine, it lands in a fiercely competitive 2026 open-weight landscape characterized by highly specialized MoE architectures. Rather than attempting to dominate every leaderboard, Thinking Machines explicitly designed Inkling—with 975 billion total and 41 billion active parameters—as a broad, balanced generalist. For example, it comes in near the middle high-end of benchmark performance 1257 on Design Arena’s Agentic Web Dev leaderboard measuring human scores of frontend web design. But China’s leading AI labs have produced models with elite reasoning and coding capabilities, posing a stiff challenge to Inkling's generalist approach and ultimately outperforming it on general and coding benchmarks. GLM 5.2: Widely considered the top open-weight reasoning model available in the benchmark set, GLM 5.2 outperforms Inkling on pure coding, agentic, and complex reasoning tasks. It scores 62.1% on SWEBench Pro (Public) compared to Inkling’s 54.3%, and a massive 82.7 on Terminal Bench 2.1 against Inkling’s 63.8. GLM 5.2 also holds the edge in text-only reasoning, scoring 40.1% on HLE (text only) versus Inkling's 30.0%. DeepSeek V4 Pro: DeepSeek maintains an edge in several strict coding and factuality domains, beating Inkling on SWEBench Verified (80.6% vs. 77.6%) and SimpleQA Verified (57.0% vs. 43.9%). However, Inkling successfully overtakes DeepSeek V4 Pro in mathematical problem-solving, achieving 97.1% on AIME 2026 compared to DeepSeek's 96.7%. Kimi K2.6: This model outpaces Inkling across multiple technical benchmarks, delivering higher scores on GPQA Diamond (91.1% vs. 87.9%), BrowseComp (83.2% vs. 77.1%), and HLE with tools (54.0% vs. 46.0%). Yet Inkling proves more resilient on general chat instruction following, scoring 79.8% on IFBench compared to Kimi K2.6's 76.0%. Against its primary U.S.-based open-weight competition, Inkling demonstrates strong parity and frequent superiority. Nemotron 3 Ultra: Inkling consistently outperforms this U.S. rival across reasoning and coding. Inkling posts 97.1% on AIME 2026 and 77.6% on SWEBench Verified, beating Nemotron's 94.2% and 70.7%, respectively. Furthermore, Inkling significantly leads in agentic workflows, scoring 74.1% on MCP Atlas against Nemotron's 44.7%. When compared to closed-source juggernauts like Claude Fable 5, GPT 5.6 Sol, and Gemini 3.1 Pro, Inkling trails in peak reasoning and software engineering autonomy, but remains highly competitive in multimodality. Coding and Reasoning: Closed models maintain a commanding lead. Claude Fable 5 (max) hits 95.0% on SWEBench Verified and 53.3% on HLE (text only), far outpacing Inkling's 77.6% and 30.0%. GPT 5.6 Sol dominates Terminal Bench 2.1 with an 89.5, easily clearing Inkling's 63.8. Native Multimodality: Inkling's native visual and audio capabilities hold their own. On the MMMU Pro (Standard 10) vision benchmark, Inkling's 73.3% is competitive, though trailing Claude Fable 5's 84.2% and GPT 5.6 Sol's 83.0%. In audio processing, Inkling scores a highly respectable 77.2% on MMAU, keeping it within striking distance of Gemini 3.1 Pro's 82.5%. If an enterprise workflow demands elite software engineering autonomy or the highest bounds of text-only reasoning, models like GLM 5.2 or proprietary systems like Claude Fable 5 maintain the edge. However, Inkling carves out a unique and highly defensible position: it is the most capable open-weight foundation model that natively fuses text, vision, and audio, while simultaneously offering developers direct programmatic control over the cost-to-performance ratio. The Shift from Static Reasoning to Controllable Thinking Rather than attempting to build a singular "god model" optimized strictly for state-of-the-art benchmark domination, Thinking Machines engineered Inkling for adaptability and efficiency in real-world workflows. The standout feature of this release is Inkling's "controllable thinking effort." Developers can programmatically adjust the model's reasoning budget—scaling from 0.2 to 0.99—to dictate how hard the AI should "think" before generating an output. As the company noted, "Inkling's continuous thinking effort lets you pick your point on the cost/performance curve—reaching the same score with a fraction of the tokens". In practical terms, this allows enterprises to deploy Inkling with lower token expenditure for simpler tasks, while cranking up the compute overhead for complex, multi-step reasoning challenges. However, by keeping the thinking effort lower and generating fewer tokens, the cost-conscious enterprise can achieve high quality results and performance on simple tasks while spending less money, or, in the case of those running models locally, less costs on energy and compute resources. During the model’s large-scale reinforcement learning (RL) training over 30 million rollouts, researchers observed an emergent phenomenon they called "chain of thought condensation". Over time, Inkling naturally learned to compress its internal reasoning steps—dropping grammatical overhead and connectives—while reaching the same accurate conclusions, resulting in drastically reduced latency. Epistemics and Censorship Resistance A notable element of Thinking Machines' release is its explicit focus on the model's epistemics—specifically its calibration, instruction following, and resistance to censorship. In an ecosystem where open-weight models adopt either overly restrictive safety guardrails or echo state-aligned ideological talking points, Inkling was intentionally trained to answer directly on politically sensitive or heavily censored topics. To validate this approach, Thinking Machines submitted Inkling to the Propaganda and Censorship Eval developed by AI startup Cognition. According to the published findings, Inkling demonstrated "strong patterns of censorship non-compliance," effectively resisting ideological capture or boilerplate refusals when presented with sensitive subjects. Despite its resistance to censorship, the model maintains a robust defense against genuinely malicious, dangerous, or illegal queries. On the StrongREJECT benchmark—which tests responses to unambiguous harmful requests—Inkling scored 98.6%, placing it in line with strict frontier safety standards. Furthermore, on the FORTRESS benchmark, Inkling successfully navigated the line between safety and over-refusal: it achieved a 78.0% refusal rate on adversarial queries (such as those involving weapons, cyberattacks, or violence) while maintaining a 95.9% compliance rate on benign, look-alike queries. Thinking Machines noted that typical open-weight vulnerabilities remain within the architecture. Internal safety evaluations revealed an "occasional tendency to comply with role-play and indirectly framed prompts concerning harmful topics". The company advised enterprise developers to treat the model's built-in refusals as just one layer of security, recommending the downstream deployment of external moderation tools—such as Llama Guard—to filter adversarial jailbreaks and enforce use-case-specific safety policies at the application level. Under the Hood: Architecture and Multimodality Inkling's scale is staggering, yet sparse. The MoE architecture features 975 billion total parameters, but only 41 billion parameters are active during any given token generation. It supports a massive context window of 1 million tokens and diverges from typical transformer models by using relative positional embeddings instead of the industry-standard Rotary Positional Embedding (RoPE). True to the company's foundational vision, Inkling was trained from scratch to be natively multimodal. Unlike models that rely on bolted-on external encoders, Inkling uses an encoder-free early fusion approach. It directly ingests audio as discrete dMel spectrograms and visual data as 40x40 pixel patches via a hierarchical multi-layer perceptron (hMLP), projecting all modalities into a shared hidden space. Licensing: True Open-Source for the Enterprise For enterprise IT teams and developers, the most disruptive aspect of Inkling may be its licensing. Inkling is released under the permissive Apache 2.0 license. In an ecosystem where many so-called "open" models from Western labs are tethered to dual-use commercial licenses, acceptable use restrictions, or revenue caps, an Apache 2.0 designation makes Inkling a true open-source foundation. This gives developers the legal freedom to download, modify, integrate, and commercialize the model weights entirely royalty-free. The model is readily deployable across major open-source inference libraries—including SGLang, vLLM, TokenSpeed, and llama.cpp—and comes with a native NVFP4 quantized checkpoint optimized for NVIDIA Blackwell systems. Community Reactions: The Engineering Feat The AI community's response has been swift, praising both the model's openness and the underlying engineering execution. In a post on X , Thinking Machines co-founder John Schulman reflected on the rapid development cycle: "Inkling is out today, with open weights and in Tinker. It's been fun to watch this one come together: pretraining began last winter, and starting in mid-January a small team built up the coding, reasoning, and agentic training from there. We learned a lot building it, and I hope people find good uses for it." Horace He, a researcher at Thinking Machines (previously from PyTorch), underscored the difficulty of the task in another post on X : "It truly takes a village to release a model, perhaps especially an open weights model. Actually doing the entire process from scratch, from data to pretraining to posttraining to actual release, gives a lot of appreciation for anyone who does it!" The broader open-source ecosystem has also embraced the technical integrations. Lysandre Debut, the Chief Open-Source Officer at Hugging Face, shared his enthusiasm regarding the model's optimization in his own X post : "One thing I find quite striking is how much easier accelerating models has become... We replaced the model's causal Conv1D with the `causal-conv1d` kernel. One line changed, +4% tokens per second. We then replaced its attention implementation with FlashAttention-4. Another single change, another +11%. That's a total throughput improvement of about 15%, without changing the model architecture or retraining anything." Tiezhen Wang, an ecosystem growth expert and ex-Googler, celebrated the release as a massive win for the open-source community, listing the model's impressive specifications on X, highlighting its "975B total, 41B active" size, "Native MTP support," and the highly coveted "Apache 2.0 license." Background: The Road to Inkling To understand the significance of Inkling, one has to look back at the rapid trajectory of Thinking Machines over the past 18 months. When Mira Murati departed OpenAI in late 2024 to found Thinking Machines alongside industry veterans like John Schulman and Barret Zoph, the stated goal was to pivot away from building isolated autonomous agents. Instead, the company aimed to build flexible, multimodal systems designed for genuine human-AI collaboration and open science. By July 2025, the startup had secured a historic $2 billion seed round led by Andreessen Horowitz at a $12 billion valuation. At the time, Murati promised the impending release of a product with a "significant open source component" to empower researchers and startups. The company’s philosophy began coming into sharper focus in October 2025 with the launch of Tinker , a Python-based API for large language model fine-tuning that gave researchers granular control over training pipelines without the friction of distributed compute management. That same month, Thinking Machines researcher Rafael Rafailov delivered a provocative critique of the AI industry at TED AI . He argued that the current trajectory of simply throwing more compute at models was fundamentally flawed, noting that today's systems take shortcuts—like wrapping code in try/except blocks—because they are trained strictly for task completion rather than genuine learning. Rafailov posited that the first artificial superintelligence would not be a "god model," but rather a "superhuman learner" capable of meta-learning and internalizing abstractions. Inkling’s architecture—specifically its controllable thinking effort and its ability to organically compress its chain of thought during RL—feels like the first tangible realization of Rafailov's thesis. In May 2026, the lab teased its technical prowess with the research preview of TML-Interaction-Small , a system that eliminated "turn-based" chat by processing inputs and outputs simultaneously in 200ms chunks. This "full-duplex" breakthrough proved the company could build highly responsive, natively multimodal models from scratch. Now, with Inkling out in the wild, Thinking Machines has delivered on its foundational promises. By offering a massive, natively multimodal model under a true open-source license, they aren't just giving developers a new tool—they are attempting to fundamentally rewrite the economics and accessibility of frontier AI development.
- Uncertainty Quantification for LLM Function-Calling
Large Language Models (LLMs) are increasingly deployed to autonomously solve real-world tasks. A key ingredient for this is the LLM Function-Calling paradigm, a widely used approach for equipping LLMs with tool-use capabilities. However, an LLM calling functions incorrectly can have severe implications, especially when their effects are irreversible, e.g., transferring money or deleting data. Hence, it is of paramount importance to consider the LLM’s confidence that a function call solves the task correctly prior to executing it. Uncertainty Quantification (UQ) methods can be used to quantify…
Score: 49🌐 MovesJul 15, 2026https://machinelearning.apple.com/research/uncertainty-quantification-function-calling - Daydreaming algorithm helps AI remember what matters
During the day, our brain acquires new memories; at night, during sleep, it consolidates the important ones and eliminates the useless ones. A similar principle has been applied to Hopfield networks, one of the classic models of artificial intelligence inspired by the workings of the brain. In 2025, Federico Ricci-Tersenghi and colleagues developed Daydreaming, an algorithm that combines the learning of new memories with the elimination of spurious ones, drastically improving the network's capacity.
- Arup partners YJK to launch AI Designer in Hong Kong to advance AI-enabled structural engineering
Arup partners YJK to launch AI Designer in Hong Kong to advance AI-enabled structural engineering
- Kerala's iHUB Robotics is teaching robots to see, act, and understand
Kerala's iHUB Robotics is teaching robots to see, act, and understand YourStory.com
- As AI becomes a gatekeeper, can India Inc influence what it recommends?
As consumers increasingly ask AI assistants what to buy instead of searching Google, India Inc faces a new challenge
- I Gave Claude a Memory That Survives Between Conversations — Here’s the MCP Server That Does It
A tested, running MCP server that gives any AI agent persistent long-term memory — it remembers facts and preferences, reinforces the ones… Continue reading on Towards AI »
- Gecko Robotics to open new manufacturing facility in Sewickley focused on defense
The 10,000-square-foot facility in Sewickley aims to integrate robotics into defense and manufacturing. Move-in is already happening.
Score: 48🌐 MovesJul 15, 2026https://www.bizjournals.com/pittsburgh/news/2026/07/15/gecko-sewickley-manufacturing-facility.html?ana=brss_6150 - Founder sells off Scottsdale software developer, AI studio
After more than 16 years of leading Founders Workshop, Vincent Serpico has sold the Valley innovation app builder. The new owner has software industry experience spanning more than 30 years.
Score: 48💰 MoneyJul 15, 2026https://www.bizjournals.com/phoenix/news/2026/07/15/scottsdale-founders-workshop-acquired.html?ana=brss_6150 - AI startup boss warns UK cannot become ‘dependent’ on overseas tech
UK AI infrastructure company Valarian has raised $50m (£37m) in fresh funding as its boss said countries should focus on controlling AI infrastructure rather than competing to build their own frontier models. The Series A round, led by Silicon Valley investor New Enterprise Associates (NEA), takes the London firm’s total funding to $70m and marks [...]
Score: 48🌐 MovesJul 15, 2026https://www.cityam.com/ai-startup-boss-warns-uk-cannot-become-dependent-on-overseas-tech/ - Google's AI Has Access to More Than You Think. Change These 7 Settings Now to Protect Your Privacy
Google's AI Has Access to More Than You Think. Change These 7 Settings Now to Protect Your Privacy PCMag UK
Score: 48🌐 MovesJul 15, 2026https://uk.pcmag.com/ai/164639/google-has-access-to-more-than-you-think-protect-privacy-when-using-gemini - Skill Discovery for Software Scripting Automation via Offline Simulations with LLMs
Skill Discovery for Software Scripting Automation via Offline Simulations with LLMs irep.mbzuai.ac.ae
Score: 48🌐 MovesJul 15, 2026https://irep.mbzuai.ac.ae/bitstreams/4d8f5588-46ac-432d-91a5-3aa623ef4429/download - Hamilton city council rejects one-year pause on AI data centres
Hamilton city council rejects one-year pause on AI data centres Toronto Star
- Business leaders seek AI-led growth at Jeju Forum
South Korea’s leading business figures gathered Wednesday on Jeju Island to seek new sources of growth as artificial intelligence reshapes the country’s industries, workplaces and education system. The Korea Chamber of Commerce and Industry opened its 49th Jeju Forum at the Shilla Jeju hotel, bringing together about 500 corporate executives, regional chamber leaders and senior government officials for four days of talks through Saturday. KCCI Chairman Chey Tae-won, who also heads SK Group, opene
- Agents need their own computer. Here's how to give them one safely.
Guide on providing agents with dedicated computing resources while ensuring safety.
- The next challenge for coding agents
The next challenge for coding agents InfoWorld
Score: 48🌐 MovesJul 15, 2026https://www.infoworld.com/article/4196780/the-next-challenge-for-coding-agents.html - CLaRa: Bridging Retrieval and Generation with Continuous Latent Reasoning
Retrieval-augmented generation (RAG) enhances large language models (LLMs) with external knowledge but still suffers from long contexts and disjoint retrieval–generation optimization. In this work, we propose CLaRa (Continuous Latent Reasoning), a unified framework that performs embedding-based compression and joint optimization in a shared continuous space. To obtain semantically rich and retrievable compressed vectors, thereby reducing the document length fed into the generator, we introduce SCP, a key-preserving data synthesis framework based on question-answering and paraphrase…
- One Layer Is Enough: Adapting Pretrained Visual Encoders for Image Generation
Visual generative models (e.g., diffusion models) typically operate in compressed latent spaces to balance training efficiency and sample quality. In parallel, there has been growing interest in leveraging high-quality pre-trained visual representations—either by aligning them inside VAEs or directly within the generative model. However, adapting such representations remains challenging due to fundamental mismatches between understanding-oriented features and generation-friendly latent spaces. Representation encoders benefit from high-dimensional latents that capture diverse hypotheses for…
Score: 48🌐 MovesJul 15, 2026https://machinelearning.apple.com/research/adapting-pretrained-visual-encoders - What Is Driving The Divergence Between AI Analyst Targets And Reality
AI stock prices are surprisingly falling, despite analysts consistently raising earnings and price targets, leading to over $2.3 trillion lost in chip stocks since June.
- KAIST develops key technology to make personalized AI safer
KAIST develops key technology to make personalized AI safer EurekAlert!
- Indonesia 'nowhere' in global AI supply chain, deputy minister says
Official says other countries are weaponizing technology chokepoints as competition shifts to hardware manufacturing and compute infrastructure.
- Seamless Remote-to-Edge AI Benchmarking: Overcoming the 3-Tier Network Bottleneck
Image by Author via AI How to bridge your GPU server, developer workstation, and local edge devices into a single-command testing pipeline. Developing deep learning models for the edge is an inherently fragmented experience. Heavy-lifting tasks — training, pruning, hardware-specific compilation (quantizing an ONNX model, compiling for an NPU) — need a beefy GPU workstation or a rented cloud instance. Execution and benchmarking, on the other hand, must happen on physical edge targets — an Android phone, a Qualcomm SoC dev board, an embedded Linux board — usually sitting on a desk somewhere else entirely. This split creates a classic three-tier network bottleneck : The AI server — remote or cloud-hosted, running the compilation toolchain. The local workstation — your everyday laptop (macOS/Windows/Linux), acting as coordinator. The edge target — an Android or Linux board reachable only via USB, Wi-Fi, or a local LAN. Figure 1: The 3-tier edge AI benchmarking topology. Diagram by author. Traditionally, developers bridge this manually: compile on the server, scp down to the laptop, plug in the target device, open an interactive adb/ssh shell, push files, run the test, pull logs back by hand. At 50 iterations a day, that manual loop burns real engineering hours — hours that should be going into the model, not the plumbing. This article walks through a zero-friction, fully automated pipeline that triggers a real on-device benchmark from your remote AI server with a single command. The Core Challenges Before writing any code, three structural pain points need addressing: Network isolation (NAT/firewalls). A remote cloud server has no direct route to a phone or micro-Linux board sitting behind your home or office router. Dynamic port allocation. Wireless debugging (ADB over Wi-Fi and similar protocols) reassigns ports on every reconnect. Hardcoded configs break constantly. Heavy transfer overhead. Re-uploading multi-gigabyte models and runtime libraries for every minor test iteration wastes bandwidth and time. Strict incremental updates are non-negotiable. Phase 1: Bypassing NAT with an SSH Reverse Tunnel The core trick is an SSH reverse tunnel : forward a port on the remote AI server back through your workstation to a device inside your local network. From the server’s point of view, the edge device becomes reachable on a local port, no inbound firewall rule required. Add a host entry to your workstation’s ~/.ssh/config: Host ai-server-remote HostName User # Forward the AI server's local port 2222 to the edge bridge's SSH port RemoteForward 2222 :8022 Here, is whatever address your edge-bridge node has on your local network — for example, an Android device running Termux with an SSH daemon on port 8022, or a small Linux hub connected to your target boards. The actual address is specific to your environment and deliberately left as a placeholder here; nothing about the pipeline depends on which private subnet you use. Once the tunnel is up, the AI server can reach localhost:2222 and land, transparently, on the edge bridge — even though the two machines were never mutually routable to begin with. Phase 2: Dynamic Target Discovery on the Edge Bridge Wireless-debugging targets shift ports on reconnect, so the pipeline needs to self-heal rather than rely on a fixed port number. A small helper script, running on the edge bridge itself , sweeps a standard debugging port range and reconnects automatically: #!/usr/bin/env bash # auto_adb_scan.sh - automated wireless ADB connection recovery # 1. Clean up stale/offline connections OFFLINE_DEVICES=$(adb devices | awk '$1 ~ /:/ && $2 != "device" {print $1}') if [ -n "$OFFLINE_DEVICES" ]; then echo "[INFO] Cleaning up offline ADB connections..." for DEV in $OFFLINE_DEVICES; do adb disconnect "$DEV" > /dev/null 2>&1 done fi # 2. Skip scanning if a device is already connected DEVICE_COUNT=$(adb devices | grep -cw "device") if [ "$DEVICE_COUNT" -gt 0 ]; then echo "[SUCCESS] Active target connected. Skipping scan." exit 0 fi echo "[INFO] No target detected. Scanning the local loopback interface..." # 3. Sweep a standard high-range debugging port window. # 127.0.0.1 here is the universal loopback address - every machine has # one, and it reveals nothing about the underlying network; it is not # an address specific to this setup. OPEN_PORTS=$(nmap -p 30000-60000 --open 127.0.0.1 2>/dev/null \ | grep -E "^[0-9]+/tcp" | awk '{print $1}' | cut -d'/' -f1) if [ -z "$OPEN_PORTS" ]; then echo "[ERROR] No open debugging interfaces found. Is wireless debugging enabled?" exit 1 fi # 4. Try each candidate port until one connects for PORT in $OPEN_PORTS; do echo "[INFO] Attempting connection on port $PORT..." RESULT=$(adb connect 127.0.0.1:$PORT 2>&1) if [[ "$RESULT" == *"connected to"* ]] || [[ "$RESULT" == *"already connected"* ]]; then echo "[SUCCESS] Connected to target at 127.0.0.1:$PORT" exit 0 fi done echo "[ERROR] No candidate port produced a working ADB session." exit 1 Note that this script runs entirely on the edge bridge’s own loopback — it never needs to know the outside world’s addressing scheme at all, which is part of why the pattern stays portable across environments. Phase 3: The Unified Incremental Execution Engine The master controller lives on the AI compilation server and enforces two design rules: Conda environment isolation. Toolchains often run inside Conda envs, which can silently inject an incompatible OpenSSL/crypto path into ssh/scp — a common, hard-to-diagnose cause of "SSH works interactively but fails in scripts." The runner forces the system binaries. Tiered file-sync strategy. Files are grouped as fixed-size binaries , version-stamped shared libraries , and raw assets , each checked with the cheapest sufficient method (MD5 for small binaries, size/version comparison for large SDK libraries), then bundled into one archive transfer instead of many small ones. #!/usr/bin/env bash # run_on_device_tests.sh — AI server -> edge target deployment engine set -e FORCE_UPLOAD=0 for arg in "$@"; do [ "$arg" == "--force" ] || [ "$arg" == "-f" ] && FORCE_UPLOAD=1 done # --- 1. Prevent Conda PATH pollution --- SYS_SSH="/usr/bin/ssh" SYS_SCP="/usr/bin/scp" [ -f "$SYS_SSH" ] || SYS_SSH=$(which -a ssh | grep -v "$CONDA_PREFIX" | head -n 1) [ -f "$SYS_SCP" ] || SYS_SCP=$(which -a scp | grep -v "$CONDA_PREFIX" | head -n 1) SSH_OPTS="-o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null -o LogLevel=ERROR" safe_ssh() { env LD_LIBRARY_PATH= "$SYS_SSH" $SSH_OPTS "$@"; } safe_scp() { env LD_LIBRARY_PATH= "$SYS_SCP" $SSH_OPTS "$@"; } # --- 2. Paths (project-specific values omitted) --- BASE_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)" REMOTE_TARGET_DIR="/data/local/tmp/edge_ai_workspace" BRIDGE_STAGING_DIR="\$HOME/staging_area" # on the edge bridge LOCAL_OUTPUT_DIR="$BASE_DIR/results/device_test" mkdir -p "$LOCAL_OUTPUT_DIR" # --- 3. Verify tunnel + auto-heal target connection --- echo "[STEP 1] Validating SSH bridge and scanning target ports..." safe_ssh edge-bridge "~/auto_adb_scan.sh" || true DEVICE_CONNECTED=$(safe_ssh edge-bridge "adb devices" | grep -cw "device") if [ "$DEVICE_CONNECTED" -eq 0 ]; then echo "[FATAL] No active edge device found. Terminating." exit 1 fi safe_ssh edge-bridge "mkdir -p $BRIDGE_STAGING_DIR" >/dev/null 2>&1 || true safe_ssh edge-bridge "adb shell 'mkdir -p $REMOTE_TARGET_DIR'" >/dev/null 2>&1 || true # --- 4. Incremental delta check --- declare -a SYNC_MANIFEST=() evaluate_delta() { local LOCAL_FILE=$1 TARGET_FILE=$2 IS_EXEC=$3 STRATEGY=${4:-md5} [ -f "$LOCAL_FILE" ] || return if [ "$FORCE_UPLOAD" -eq 1 ]; then SYNC_MANIFEST+=("$LOCAL_FILE|$TARGET_FILE|$IS_EXEC"); return fi if ! safe_ssh edge-bridge "adb shell '[ -f $TARGET_FILE ]'" >/dev/null 2>&1; then SYNC_MANIFEST+=("$LOCAL_FILE|$TARGET_FILE|$IS_EXEC"); return fi if [ "$STRATEGY" == "size" ]; then local L=$(stat -c%s "$LOCAL_FILE" 2>/dev/null || stat -f%z "$LOCAL_FILE") local R=$(safe_ssh edge-bridge "adb shell 'stat -c%s $TARGET_FILE'" | tr -d '\r' | tail -n1) [ "$L" != "$R" ] && SYNC_MANIFEST+=("$LOCAL_FILE|$TARGET_FILE|$IS_EXEC") else local L=$(md5sum "$LOCAL_FILE" 2>/dev/null | awk '{print $1}') local R=$(safe_ssh edge-bridge "adb shell 'md5sum $TARGET_FILE'" | awk '{print $1}') [ "$L" != "$R" ] && SYNC_MANIFEST+=("$LOCAL_FILE|$TARGET_FILE|$IS_EXEC") fi } evaluate_delta "$BASE_DIR/build/bin/inference_runner" \ "$REMOTE_TARGET_DIR/inference_runner" 1 "md5" evaluate_delta "$BASE_DIR/models/quantized_backbone.onnx" \ "$REMOTE_TARGET_DIR/quantized_backbone.onnx" 0 "size" evaluate_delta "$BASE_DIR/assets/sample_16k.wav" \ "$REMOTE_TARGET_DIR/sample_16k.wav" 0 "size" # --- 5. Package, upload, deploy only what changed --- if [ ${#SYNC_MANIFEST[@]} -gt 0 ]; then echo "[STEP 2] Syncing ${#SYNC_MANIFEST[@]} changed file(s)..." TMP=$(mktemp -d) for entry in "${SYNC_MANIFEST[@]}"; do IFS='|' read -r L R E <<< "$entry"; cp "$L" "$TMP/" done tar -czf "$TMP/payload_delta.tar.gz" -C "$TMP" . >/dev/null safe_scp "$TMP/payload_delta.tar.gz" edge-bridge:$BRIDGE_STAGING_DIR/ safe_ssh edge-bridge "tar -xzf $BRIDGE_STAGING_DIR/payload_delta.tar.gz -C $BRIDGE_STAGING_DIR/" for entry in "${SYNC_MANIFEST[@]}"; do IFS='|' read -r L R E <<< "$entry" F=$(basename "$L") safe_ssh edge-bridge "adb push $BRIDGE_STAGING_DIR/$F $R" [ "$E" == "1" ] && safe_ssh edge-bridge "adb shell 'chmod +x $R'" done rm -rf "$TMP" safe_ssh edge-bridge "rm -rf $BRIDGE_STAGING_DIR/*" else echo "[STEP 2] All target assets already up to date." fi # --- 6. Run benchmark, pull results back through the same tunnel --- echo "[STEP 3] Executing remote benchmark..." RUN_CMD="cd $REMOTE_TARGET_DIR && LD_LIBRARY_PATH=$REMOTE_TARGET_DIR ./inference_runner ./quantized_backbone.onnx ./sample_16k.wav ./output_metrics.csv" safe_ssh edge-bridge "adb shell '$RUN_CMD'" > "$LOCAL_OUTPUT_DIR/run.log" 2>&1 echo "[STEP 4] Retrieving metrics..." safe_ssh edge-bridge "adb pull $REMOTE_TARGET_DIR/output_metrics.csv $BRIDGE_STAGING_DIR/" safe_scp edge-bridge:$BRIDGE_STAGING_DIR/output_metrics.csv "$LOCAL_OUTPUT_DIR/" safe_ssh edge-bridge "rm -f $BRIDGE_STAGING_DIR/output_metrics.csv" echo "[SUCCESS] Cycle complete → $LOCAL_OUTPUT_DIR/output_metrics.csv" ( edge-bridge above is an SSH config alias, not a literal hostname — the same pattern used for ai-server-remote in Phase 1. Swap in your own ~/.ssh/config entries; nothing else in the script needs to change.) The full request/response cycle, end to end, looks like this: Figure 2: One pipeline cycle, from trigger to results — no manual step in between. Diagram by author. Why This Architecture Works 1. Zero-friction developer experience. Once configured, the entire round trip — port discovery, delta upload, execution, result retrieval — runs from one terminal command on the AI server, whether you’re triggering it from a plain shell or from an IDE’s integrated terminal over SSH. 2. Bandwidth discipline. Comparing MD5/size before every transfer, then batching whatever did change into a single .tar.gz, avoids both wasted re-uploads and the round-trip overhead of many small scp calls — the two biggest hidden costs in naive device-testing scripts. 3. Topology independence. The AI server never talks to a raw IP address; it always talks to an SSH config alias. Whether the edge bridge is an Android phone on your desk or a Linux hub three buildings away, only the ~/.ssh/config entry changes — the script doesn't. Illustrative Impact The table below shows the shape of the improvement you should expect from adding auto-discovery and incremental sync to a manual workflow — exact numbers will vary with your network, model size, and device: Stage Manual Workflow Automated Pipeline (first run) Automated Pipeline (incremental) Model verification Manual eyeballing Automated (MD5/size check) Automated (fast skip) Network/device discovery Manual cable/adb fumbling Automated port sweep Cached, near-instant Transport scp → local → adb push Single compressed tunnel transfer Skipped if unchanged Relative cycle time Slowest (minutes, manual) Fast (full transfer, once) Fastest (seconds, delta only) Table 1: Qualitative comparison of workflow stages, not measured benchmark data. Table by author. Submission Notes and Attribution All diagrams (Figures 1–2) and the comparison table (Table 1) in this article were created by the author specifically for this piece; no third-party images, charts, or data were used. The code shown here is a simplified, generalized rewrite illustrating the architecture pattern; it is provided for educational reference rather than as a drop-in, license-free package — adapt it to your own project’s licensing terms before reuse. No real hostnames, IP addresses, or credentials from any production environment appear in this article; all addresses and paths are placeholders (<...> or generic aliases like ai-server-remote / edge-bridge). Closing Thought Restructure your development path around a three-tier SSH tunnel plus a smart incremental-sync engine, and the physical gap between “compiled in the cloud” and “verified on the actual chip” stops being a manual chore — it becomes a single command you can run fifty times a day without thinking about it. Happy edge hacking. If you found this article helpful, feel free to connect with me on LinkedIn: https://www.linkedin.com/in/cobengao Seamless Remote-to-Edge AI Benchmarking: Overcoming the 3-Tier Network Bottleneck was originally published in Towards AI on Medium, where people are continuing the conversation by highlighting and responding to this story.
- Shakti Pumps partners Salesforce to build AI-first agri-tech platform
Shakti Pumps partners Salesforce to build AI-first agri-tech platform Techcircle
Score: 47🌐 MovesJul 15, 2026https://www.techcircle.in/2026/07/15/shakti-pumps-partners-salesforce-to-build-ai-first-agri-tech-platform - [CVPR 2026] Beyond Text: How VinQA Incorporates Visual Elements into Responses for Multimodal Document QA
[CVPR 2026] Beyond Text: How VinQA Incorporates Visual Elements into Responses for Multimodal Document QA
- Towards demystifying the creativity of diffusion models
Algorithms & Theory
Score: 47🌐 MovesJul 15, 2026https://research.google/blog/towards-demystifying-the-creativity-of-diffusion-models/ - Scalable Oversight Across Generative Visual AI: Toward Visual Storytelling for Everyone
Scalable Oversight Across Generative Visual AI: Toward Visual Storytelling for Everyone Carnegie Mellon University
- Fynd launches 'Fynd Create' in GCC, an AI-native platform for fashion design, sourcing and production
Early deployments across global fashion brands have demonstrated up to 60% improvements in design productivity
- Albright College to Offer AI Training Program for Teachers
A private liberal-arts college in Reading, Pa., is using a grant from a STEM-focused foundation to launch a fully online, four-course AI Endorsement program to prepare teachers to integrate AI into their classrooms.
Score: 46🌐 MovesJul 15, 2026https://www.govtech.com/education/higher-ed/albright-college-to-offer-ai-training-program-for-teachers - Suno now lets iPhone users generate songs directly in iMessage
Starting today, iPhone users who have the Suno app installed can generate songs from the Messages app using text or voice prompts. Here are the details.
Score: 46🌐 MovesJul 15, 2026https://9to5mac.com/2026/07/15/suno-now-lets-iphone-users-generate-songs-directly-in-imessage/ - The hidden AI cost driver: Harness design can make or break enterprise agent economics
A largely overlooked layer of the AI stack is emerging as a major driver of enterprise costs. New testing by AI consultancy Systima found that agent harnesses, the software that coordinates models, tools and workflows, can generate significant token overhead through their configuration alone, potentially inflating the cost of AI deployments as organizations scale agents from experimental pilots to production environments. The firm, which ran a series of tests by juxtaposing two harnesses on the same tasks, namely Anthropic’s Claude Code and open-source OpenCode using the same Claude Sonnet 4.5 model underneath, found both exhibiting sharply different token overhead because of the differences in their configuration. These differences included system prompts, tool definitions, agent coordination mechanisms and other orchestration components, resulting in markedly different baseline input token overhead before users even entered a prompt, the consultancy firm wrote in a blog post . Separately, the firm also found that other configuration choices while setting up the harnesses such as repository instruction files, Model Context Protocol (MCP) servers, prompt framework templates and subagents can each add substantial token overhead. The consultancy’s conclusions are also supported by emerging academic research examining how orchestration of the harnesses themselves, rather than optimizing models or changing them, can help enterprises reshape the economics around AI agents. In a paper , titled The Harness Effect: How Orchestration Design Sets the Token Economics of Enterprise Agentic AI, researchers showed that changing the harness while keeping models and tasks the same can reduce token consumption by 38%, cost per task by 41%, and execution time by 44% while maintaining comparable quality. Why enterprises overlook harness costs Analysts say that enterprises can gain greater control over AI agent operating costs by paying closer attention to how their harnesses are configured and orchestrated, instead of just relying on model pricing as a yardstick. “The evaluation shows that the model is only one part of agent economics. The harness, tool schemas, instructions, MCP connections, and subagents matter as well. Enterprises therefore need to measure the entire agent configuration, not assume model pricing tells them what an agent will cost,” said Stephanie Walter , practice lead of the AI stack at HyperFRAME Research. Currently, most enterprises pick agent tooling based on model quality, benchmarks, developer experience, and headline pricing per seat or per million tokens, with almost no one measuring what the harness sends per request, how stable the cache prefix is, or what subagent fan out costs at scale, echoed Advait Patel , site reliability engineer at Broadcom. “Ask the average CIO whether their coding agent rewrites its cache mid-session, and you will get a blank stare,” Patel added. However, Ashish Chaturvedi, executive research leader at HFS Research, pointed out that lack of visibility is less a failure of enterprise leaders than a consequence of how AI agent ecosystem components are sold, stacked, and managed presently. “Most organizations have no visibility, mainly due to the absence of any metric from the vendor’s end that lets CIOs measure the entire agent or at least the harness configuration. None of this shows up in the developer’s experience. The agent just works, and the tokens burn silently in the background,” Chaturvedi said. The problem is further compounded, according to Chaturvedi, due to the manner in which AI agent configuration is distributed across enterprise teams. “The harness is chosen by one team, the instruction file written by another, and the MCP servers attached by a third, so no single person sees the cumulative weight,” Chaturvedi noted. Even when, in some cases, enterprises do have visibility and ownership, Patel argued, the industry, in general, still lack the operational maturity and discipline to systematically optimize AI agent costs. “FinOps for agents is where cloud FinOps was in 2013. Nobody has hired the equivalent of a cost optimization team focused on prompt engineering, harness configuration, and cache stability,” Patel said. Separately, Abhishek Satapathy , principal analyst at Avasant, pointed out that the invisibility issue stems from how enterprises evaluate AI agents before deploying them into production: “Most proof-of-concepts involve a limited number of users, relatively short-lived sessions, and controlled agentic interactions, where the accuracy of model output is the primary evaluation criterion.” The analysts’ comments also echo the conclusions of another research paper , in which researchers argued that token consumption in agentic software engineering systems remains poorly understood because existing metrics provide limited visibility into where tokens are spent across orchestration components. How CIOs can improve visibility into AI agent costs Closing that visibility gap, though, according to Satapathy, is increasingly becoming a priority for enterprises, as AI agents move from pilots to production and operating costs become harder to predict. “Across our advisory engagements, we are seeing growing demand for AI observability frameworks that combine runtime tracing, workload-level cost attribution, and execution analytics. This enables organizations to establish engineering baselines, benchmark workload efficiency, forecast AI operating costs, and continuously optimize agent performance as deployments mature,” Satapathy said. However, until vendors provide more comprehensive visibility into harness-level token consumption, analysts said enterprises should begin treating harness configuration as an operational governance issue rather than merely a developer preference. “The single most valuable move is to get visibility into what the harness actually sends. Enterprises should treat configuration as a governed cost decision, deliberately match harnesses to workloads, and closely monitor cache behavior and subagent fan-out, since those were among the biggest cost multipliers identified in the evaluation,” Chaturvedi said. Walter echoed that recommendation, saying CIOs should require observability across the entire agent configuration: “Without that visibility, enterprises are effectively buying an agent platform without knowing how much of the bill comes from useful work versus orchestration overhead.”
- Perplexity launches secure sandbox to make its AI agents secure and powerful
Perplexity AI Inc. today introduced a new feature that takes its current agentic artificial intelligence service, Computer, to perform better with greater security. The company introduced SPACE, a sandbox platform designed to allow its AI agent to act with its full capabilities, while providing the highest level of security for agentic systems. Perplexity Computer can […] The post Perplexity launches secure sandbox to make its AI agents secure and powerful appeared first on SiliconANGLE .
Score: 46🌐 MovesJul 15, 2026https://siliconangle.com/2026/07/15/perplexity-launches-secure-sandbox-make-ai-agents-secure-powerful/ - New York offers Democrats a roadmap for data center fight
New York Gov. Kathy Hochul's first-in-the-nation data center moratorium could provide a playbook for other Democrats confronting one of the midterms' most heated issues. Why it matters: Voters across the country are protesting data centers and looking for politicians to take decisive action to address concerns over the industry's rapid expansion. Until now, no governor had imposed a statewide data center moratorium. Hochul's approach — using executive authority while continuing to negotiate over broader legislation — could become a playbook for other governors. Zoom in: Hochul went the executive order route in order to move as quickly as possible, her staffers say, while continuing to work with the New York legislature. State Sen. Kristen Gonzalez recently pushed a data center moratorium bill through the New York legislature, but instead of signing that, Hochul took executive action. Hochul's staffers said that Gonzalez's legislation raised more complicated policy questions that require further negotiation. For example, Gonzalez said, negotiations will likely include clearer guidelines for public hearings. Gonzalez told Axios that she's "really supportive" of Hochul's approach and worked with the governor's office because she agreed the state needed to act quickly. The fine print: Gonzalez's bill goes further than Hochul's order. It would impact projects above the 20 megawatt level instead of 50, create new electric and water rate classes and require public hearings before future permits are approved. Hochul's staffers say the governor wants to pair her executive order with legislation to ensure hyperscalers don't get tax benefits, while continuing to negotiate with Gonzalez on her bill. Moratorium bills are cropping up across the country. Gonzalez said this will "certainly support" other governors considering proposals that are likely to make it to their desks. Zoom out: The backlash isn't limited to blue states. Texas Gov. Greg Abbott (R) recently called for banning new AI data centers in rural neighborhoods and to require the industry to shoulder more of its infrastructure costs. Between the lines: Hochul is toeing the line between companies and voters, and acting fast without shying away from language viewed by some other Democrats as too politically charged. "Now, no one can accuse New York of fearing innovation," Hochul said at a signing ceremony on Tuesday. "But that said, a lot of Americans — for a lot of them, the specter of unchecked AI brings up fear, anxiety, a lot of worry." Hochul and President Trump have aligned around efforts to have companies pay for the energy they demand. The intrigue: Sen. Bernie Sanders (I-Vt.) and Rep. Alexandria Ocasio-Cortez (D-N.Y.) thrust the idea of a data center moratorium into the national spotlight earlier this year with legislation that would have paused new projects nationwide. Sanders on Tuesday noted the idea, once "radical," is becoming a reality in places like New York. The Sanders-AOC proposal went even further than Gonzalez's bill, and would lift the ban only after Congress passed sweeping AI safety legislation covering issues from civil rights to consumer protection. The bottom line: Whether New York proves to be an outlier or becomes a model, Hochul has given governors a new playbook for responding to the data center backlash.
Score: 45🌐 MovesJul 15, 2026https://www.axios.com/2026/07/15/new-york-democrats-roadmap-data-center-fight - OpenAIs smart speaker sounds like a cross between a HomePod and a Furby
OpenAI is reportedly making a portable, screenless smart speaker intended to act as a household AI companion.
- The Unit Economics Of India’s Voice AI Boom
India’s voice AI revolution has revolved around a few key pillars, such as speech recognition accuracy, multilingual capabilities, latency and…
- Google Releases LiteRT.js: A JavaScript Binding of LiteRT That Runs .tflite Models in Browsers via WebGPU
Google Releases LiteRT.js: A JavaScript Binding of LiteRT That Runs .tflite Models in Browsers via WebGPU MarkTechPost