AI News Archive: August 14, 2026 — Part 6
Sourced from 500+ daily AI sources, scored by relevance.
- 采用OaAS模式，克服AI资源有限、缺乏成功经验与投资回报率不确定问题
采用OaAS模式，克服AI资源有限、缺乏成功经验与投资回报率不确定问题 Gartner
- Can text-native web fonts protect websites from AI companies scraping data?
Open-source project ShieldFont changes the text that basic AI scrapers extract while preserving what people see, offering websites another defence against unauthorised data collection
- ‘Enablers’ are the AI sweet spot for investors
With demand for the technology far outstripping supply, infrastructure providers will be highly sought-after
- TransFi Launches JARVIS, an AI-Powered Compliance Intelligence Platform for Cross-Border Payments
TransFi Launches JARVIS, an AI-Powered Compliance Intelligence Platform for Cross-Border Payments markets.businessinsider.com
- Before buying the Lakers for $12.5 billion, Joshua Kushner made a $1.3 billion bet on OpenAI
Before buying the Lakers for $12.5 billion, Joshua Kushner made a $1.3 billion bet on OpenAI Fortune
- Opinion | Want Concierge Medicine? AI Can Deliver It
It can analyze years of a patient’s history and flag risks in a way a physician can’t during a brief visit.
Score: 38🌐 MovesAug 14, 2026https://www.wsj.com/opinion/want-concierge-medicine-ai-can-deliver-it-102e731a?mod=rss_Technology - The Next Big Influencer Is This 4-Foot-Tall Robot From China
The Unitree G1 has found online fame as a relatively affordable robot that can charm a crowd. But can it ever hold down a real job?
- Pusan National University develops an AI framework for smarter ship navigation
Pusan National University develops an AI framework for smarter ship navigation EurekAlert!
- CVG partners with Dayton college to test drone technology and aerial systems
Sinclair has partnered with CVG airport to develop a form of aircraft with no pilots. See the details of the partnership.
- Convergence of soft electronics and artificial intelligence: from materials to intelligent systems
Convergence of soft electronics and artificial intelligence: from materials to intelligent systems EurekAlert!
- GitKraken Introduces GitLens 19, Bringing Human and AI Workflows Together in One Workbench
GitKraken Introduces GitLens 19, Bringing Human and AI Workflows Together in One Workbench USA Today
- AI時代のITスキル体系：何を保護／排除／転換すべきか
AI時代のITスキル体系：何を保護／排除／転換すべきか Gartner
- Quote of the day by roboticist Rodney Brooks: 'The visual appearance of a robot makes a promise about what it can do and how smart it is' — a warning about the misleading form of a humanoid robot
Just because engineers are building humanoid robots to look like us, that doesn't mean they're intelligent or even capable
- Google Meet can take notes for your in-person meetings now - here's how it works
Google's Gemini-driven meeting software can now save a transcript, send it to Google Drive, and email you a copy.
Score: 36🌐 MovesAug 14, 2026https://www.zdnet.com/article/google-meet-take-notes-in-person-meetings-how-it-works/ - Ecosystem Roundup: Why “Know Your Agent” will matter as much as KYC in payments
As AI agents start searching for suppliers, negotiating terms, and initiating payments on a company’s behalf, the old assumption that a human approves every transaction is breaking down. A new report from Sunrate and Mastercard, Beyond Automation: Defining Agentic Global Payments, argues that speed alone will not make this shift viable — businesses will need […] The post Ecosystem Roundup: Why “Know Your Agent” will matter as much as KYC in payments appeared first on e27 .
Score: 36🌐 MovesAug 14, 2026https://e27.co/ecosystem-roundup-why-know-your-agent-will-matter-as-much-as-kyc-in-payments-20260814/ - Diginex and Resulticks Sign Amended Definitive Agreement to Create a Global AI-Powered Group Spanning Customer Engagement and Trusted Sustainability Data
Diginex and Resulticks Sign Amended Definitive Agreement to Create a Global AI-Powered Group Spanning Customer Engagement and Trusted Sustainability Data markets.businessinsider.com
- How the American Executive Could Control AI Companies
Some of the most notable American AI policies to date have been enacted by unilateral executive branch action. Consider the Department of Defense’s spat with Anthropic, and the resulting threats from Pete Hegseth to invoke the Defense Production Act (DPA) against them. Or the fleeting export controls on Claude Fable/Mythos 5, manifested as a vaguely worded, threatening letter from Howard Lutnick, which might not have been legally sound but were effective anyway. The executive branch of the United States government has numerous powers that can be used to unilaterally control AI companies. We think the US executive is likely to remain heavily involved in AI governance, because the national security and foreign policy narratives about AI that empower the executive will endure. Additionally, if AI progresses very quickly, the executive will be further emboldened because it is particularly quick to respond and often entrusted with crisis management. In instances where the executive acts beyond its lawful powers, we think checks from Congress and the courts will be unreliable in restraining the executive. In this post, we: Identify and explain particular federal statutes and laws that permit the executive to act unilaterally in ways that influence—if not directly control—US AI companies. We examine: Executive powers over domestic goods and resources. Acts like the DPA grant the president broad but vague authority. Executive powers over foreign transactions. Here, the law is clearer in delegations of executive power, but has been tested in cases like Learning Resources, Inc. v. Trump. Demonstrate why making threats (“bullying”) can be an effective strategy for the executive to influence US AI companies. Examine the relationship of Congress and courts to executive power. Congressional approval is required for actions like forcibly nationalizing US AI companies or infrastructure; this acts as a meaningful limitation on some actions. However, we find that inter-branch checks on executive power are likely unreliable in instances where the executive acts unlawfully. Our aim in writing this is to illustrate the specifics of American executive dominance: in which domains the executive is powerful, where certain ambiguities and possibilities for constraint lie, and where they can’t act alone. Thank you to Alex Mallen, Buck Shlegeris, Nick Marsh, Jackson Sipple, and Cody Rushing for feedback on earlier drafts. Executive power over goods and resources related to the AI industry Presidential authority over domestic AI companies operates at the level of the goods and intellectual property produced and assets held by these companies. To exert control at this layer, the president could: Use Title I of the Defense Production Act (DPA), which forces companies to prioritize pre-existing government contracts. For example, if the Department of Defense has an existing contract with an AI company to make weapons systems, but the company has signed other contracts promising non-government actors to deliver a product by a certain date, the DPA would force the company to forgo the private contract to fulfill the government one on time. However, it’s unclear whether Title I can be used to force companies to take new contracts , like Hegseth threatened Anthropic with in early 2026. Legal discourse highlights how the language in the Act itself is ambiguous; regulations implementing this part of the DPA require accepting new contracts, but also make exemptions if the company does not usually make that product. Use Title I of DPA to restrict or allocate critical domestic materials and facilities (e.g. compute, chips). This does not require a government contract: the Title gives the executive authority to prohibit certain sales or direct resources to particular buyers without the government itself being the buyer. For example, the president might order Google to use its GPUs to serve OpenAI’s models. It is unclear if this order would be lawful. Since the Cold War, allocation authorities have rarely been used ; there is little precedent to determine the lawful scope of allocation authorities. Use Title I Section 103 of the DPA to fine or imprison individuals who interfere with the fulfillment of Title I DPA contracts. Use Title VII Section 705 of the DPA to obtain information relevant to administering or enforcing the DPA. This creates a possible mechanism for the executive to require specific evaluations or information about an AI company’s models or practices that could embolden other decisions. Biden’s EO 14110 drew in part from Section 705 in an attempt to establish a routine frontier AI company reporting scheme, but this never manifested and the order was eventually rescinded under Trump. Use the Invention Secrecy Act to block the publication, patenting, or production of research on AI capabilities to prevent other actors (within or outside the US) from catching up. The USG would maintain access to these “secrets.” Which options are most likely to be successfully used for AI? The executive is very likely to succeed in prioritizing contracts for models, hardware, and infrastructure. Prioritized contracts are extremely common. The DoD issues an estimated 300,000 priority-rated (DO and DX ratings) contracts per year. While the number of top-priority (DX-rated) contracts is much smaller, there are no functional restrictions or caps on assigning this rating besides requiring approval from the Secretary of Defense. Conversely, presidential authorities to force new contracts or allocate critical domestic materials are much more ambiguous. We generally see the resolution of ambiguities as context-dependent. While some presidents operate with more restraint, others might see ambiguities opportunistically and work to exploit them. A crisis scenario that requires urgent response (e.g. a large cyberattack on US infrastructure) likely prompts the executive to test ambiguous mechanisms. This would also bolster public-interest and national security narratives that executive authority is empowered by. We are uncertain how and if this ambiguity will be resolved, but we think it’s moderately likely the limits of executive authority will be tested in the near future. [1] Prioritized contracts could have two notable spillover effects: Fulfilling prioritized orders across a supply chain can generate soft allocation authority because prioritization can flow through manufacturers. For example, if a data center construction company has a DO-rated order, it can place DO-rated contracts on other companies that supply its parts (e.g. GPU manufacturers). This reallocates resources to a particular need specified by the government without having to invoke the more ambiguous direct allocation authority. If these contracts require specific development or deployment practices that can’t easily be applied to only government models, like pretraining only on vetted data, AI companies might adopt those practices for all of their models. Much like export controls can have domestic spillover effects that act pseudo-legislatively (e.g. effectively “banning” Fable), these contracts may provoke company-wide changes that then apply to non-contract business engagements. Finally, we think the DPA’s information authorities and the Invention Secrecy Act are presently less likely to be used for two reasons: There’s ambiguity in the language of Section 705, which potentially implies that information-related authorities can only be activated in service of a pre-existing use of the DPA (e.g. a rated contract). [2] For example, legal critiques of Biden’s executive order identified that relying on Section 705 to enforce a periodic reporting scheme was likely unlawful . However, if AI progress moves faster, or a warning shot is sufficient to prompt action, it’s possible the executive will try these authorities again over attempting to pass pre-deployment evaluation legislation or similar information laws. Many AI technologies are already patented (e.g. chips, EUV), and these can’t be unilaterally revoked. Frontier labs don’t typically produce patented products and instead just develop models with in-house techniques, so it’s unclear what the executive would actually control under the Invention Secrecy Act. Overall, we think the executive’s powers to redirect resources and command the rapid acquisition of products are highly relevant to influencing AI companies and related service providers. These powers could be used to consolidate AI production within one champion company, or get access to advanced AI technology that companies hesitate to sell. However, the outer limits of these powers remain uncertain. Executive power over foreign transactions can impact domestic AI companies Given broad foreign affairs powers, the president has various authorities that directly target transactions involving a foreign entity. Using these powers often affects domestic companies too. To control foreign transactions, the president could: Invoke the National Emergencies Act to gain a huge expanse of executive power. This includes the International Emergency Economic Powers Act , which permits the president to block transactions, sanction foreign actors, and freeze foreign assets. This would prevent other countries from investing in US AI companies in order to ensure they do not benefit from or interfere with their activities. Use export controls on inputs (chips, energy) and outputs (models, weights) to block other countries from accessing the AI supply chain. The Foreign Direct Product Rule extends American export controls to certain foreign-produced items that are the direct product of US-origin technology or software, or that are produced by equipment which is itself a direct product. For example, TSMC is subject to American export controls because their fabs run on tools that are themselves the output of American technology. Direct the Committee on Foreign Investment in the United States (CFIUS) to intervene in foreign investment relevant to US AI companies. CFIUS has executive power to investigate and attach strict conditions to foreign transactions such as mandatory board oversight. The president has delegated authority to suspend certain transactions and compel foreign divesture from American corporations. Beyond simply disrupting foreign AI development and supply chains, many of these actions have effects at the domestic and product level. For example, the Mythos and Fable export controls didn’t technically ban Anthropic from producing or deploying either model. However, deemed export laws meant foreign nationals within the US were also subject to export controls (though in what way was unclear, because the export controls were never litigated). Anthropic removed access anyway, achieving the same effect as a ban on these models. Doing a direct ban would require Congress to pass legislation, but “deemed export” rules may continue to work as an executive-only path equivalent to administering a ban. These maneuvers are useful in the instance that the executive wants to disrupt the productive capacity of American AI companies, whether for reasons of safety or belligerence. We think it is very likely that the executive uses these powers to fulfill AI governance objectives through the market. If the domestic mechanisms in the previous section are blocked or frustrated, the collateral effects of governing international transactions serve as an alternative pathway to controlling US AI companies. The executive might make threats to coerce actions it can’t directly elicit The executive could threaten the use of any of the mechanisms specified above to coerce companies into changing their actions in domains it can’t directly control. When contract negotiations between Anthropic and the Department of Defense (DoD) broke down, Pete Hegseth threatened to designate Anthropic as a supply chain risk and invoke the DPA to override Anthropic’s red lines on autonomous weapons (despite, as previously discussed, this threat’s questionable credibility). Similar threats might be used to coerce compliance with other standards. These threats are credible if the executive can legally follow through on the threat. For many potential goals, like changing corporate policy, the executive could lawfully make a threat and use a national-security pretext to justify actions that follow through on it. As we analyze later, courts are generally deferential to such pretexts. The executive can’t use its authorities to coerce individuals to give up constitutional rights or retaliate against the exercise of constitutional rights (even if those authorities are lawful in a vacuum). Threats leveraged against a company’s exercise of constitutional rights would therefore not be credible. For example, the executive can’t use threats to obtain private property, which is protected in the Fifth Amendment. SCOTUS held in Koontz v. St. Johns River Water Management District that officials can't use permitting leverage to extract concessions bearing no essential nexus or rough proportionality to the impacts of the proposed use, including when the demand is for money and when the permit is denied outright. However, unless companies sue the executive for unlawful actions, courts can’t intervene. Even when a company is confident it would win in court, relief might not be granted quickly, so the immediate cost of executive retaliation might be enough to make it give in. The Trump administration filed executive orders revoking security clearances and terminating contracts with law firms accused of targeting administration officials and using DEI policies to enable discriminatory hiring processes. While some firms went to court , Paul, Weiss instead struck a deal with the administration. Similarly, the Trump administration withheld Biden-era investment until Intel agreed to give the government a 9.9% share . It’s hard to predict if this threat would be effective in coercing companies. During a fast takeoff, the leading AI company might be vital to the US’s national security strategies, which could make the president unwilling to follow through with destructive actions. But it’s plausible the president could compel at least minor changes in corporate policy. If multiple AI companies have comparable products, then threats are more credible, since the US government has alternative suppliers for security-critical AI technologies. Inter-branch constraints on executive power are weak In this section, we describe how Congressional and judicial checks on executive power work, and why they might be weak at a structural or legal level. However, the actual effect of checks and constraints is often contingent on circumstance: the individual voting behaviour of Congress members, or the particular circuit that takes a court case. This means we cannot state that any check will certainly work or certainly fail. We are overall pessimistic about the efficacy of these constraints. The judiciary may be permissive in matters of AI governance In this post, we’ve identified areas where the president or executive is not authorized to act. However, we acknowledge that courts are limited in their capacity—and possibly their incentives—to constrain unlawful presidential actions. We are ultimately uncertain and somewhat pessimistic about the ability of judicial checks to stop unlawful executive actions. We think that national security narratives related to AI are likely to create a degree of judicial permissiveness with regard to unilateral executive action. Passivity Courts do not intervene on their own initiative: a case must reach them with evidence of demonstrable or impending harm. If affected parties don’t challenge actions, even if they are overtly illegal, courts cannot act. Further, Congress has power over the court’s jurisdiction: it can pass legislation that restricts judicial review of certain federal executive or legislative actions. While mechanisms like injunctions can be used by courts to pause actions while their lawfulness is being adjudicated, they must be persuaded to do so. Especially in matters regarding AI, it is unclear whether injunctions will be enacted against a government body. The court empowers the executive in national security The “national interest” has been invoked by the executive to justify everything from starting wars to using the DPA for baby formula . Actions framed by the executive as in the national interest, or critical for matters of national security, are generally granted broad leeway by the courts. The June 2026 Executive Order states frames its “America First cybersecurity effort” as one that “enhances both our national security and our global AI dominance.” National security narratives about AI mean the executive is empowered to govern AI, including by the courts themselves. AI-related national security efforts operate in the domain of foreign policy: export controls, military accumulation, and suspicion of foreign entities. At the same time, the executive blurs the distinction between internal and foreign affairs by asserting that certain domestic policies act in service of national security interests. This logic has already been used by the Justice Department , when it intervened to argue in favour of throwing out a case against xAI’s Colossus 2 natural gas turbines in federal court. Courts have given the executive substantial leeway during true national crises. During World War I, the executive was allowed to egregiously violate free speech protections. SCOTUS upheld race-based internment based on a fabricated military-necessity justification in World War II. If AI progress poses an existential risk to the US, it is plausible courts will uphold unconstitutional actions or avoid ruling on the merits of a case, as they did during the American Civil War with habeas c orpus cases like Ex parte Vallandigham . Beyond national security, courts can actively empower the executive— recent SCOTUS rulings explicitly delegate more power to the president alone. Failed enforcement of court rulings Lastly, the executive might ignore or slow-walk court rulings. Since courts don’t have an independent enforcement arm, there is little recourse if their rulings are disregarded. Already, the Trump administration has repeatedly defied lower court rulings on immigration with little consequence. SCOTUS rulings will probably carry more political weight than lower courts, but the executive might still find ways to stall enforcing these rulings, as they did with Kilmar Abrego Garcia . Congress controls money and legislation There are two relevant sets of constraints Congress holds: over government spending, and over the legislation that ultimately delegates the executive its power. Nationalization and appropriation require congressional approval Congress controls government finances, and could refuse to fund executive decisions. Individual committees draft spending bills to determine program and project funding. Still, many forms of executive action, like implementing export controls or prioritizing contracts, don’t require restrictively large sums of money, and are financed through an agency’s standard operating budget. For example, the DoD already has designated DPA Title III funds that it regularly uses . In the last five fiscal years (2020-2025) DPA Title III funds totaled a modest $4.4 billion USD, though the DoD is seeking an increase to approximately $30 billion for the 2027 fiscal year. Contracts and export controls might not produce the executive’s desired amount of influence or outcomes. One way of gaining a lot of control over AI, then, is directly taking over an AI company; this requires Congressional approval. Authority related nationalization requires a delegating act from Congress (like for airport security in 2001 ). These are typically specific. They can permit the seizure of property for a particular project or narrow scope of projects, like how the Department of Homeland Security only has eminent domain authority when relevant to border infrastructure and building facilities. The executive doesn’t have the authority to unilaterally nationalize an AI company or take similar actions, like replacing its board or forcibly taking a controlling equity stake. The government is constitutionally obligated to offer just compensation under the Fifth Amendment. Determining “fair market value” for appropriated property can involve the courts if the company is dissatisfied with the government’s offer. Here, it could likely be difficult to establish “fair market value” for newer, less tangible goods like model weights. The lack of precedent means it is more likely for these appropriation attempts to manifest as disputes in court. The high likelihood of other branches of government being drawn into matters of AI-related appropriation means it is difficult for the executive to act unilaterally in appropriating or nationalizing AI. Another limiting factor is the sheer cost of doing so. AI company acquisitions would likely cost in the magnitude of hundreds of billions, if not trillions. While smaller appropriations of a portion of a company’s assets (e.g. a data center site through eminent domain) or purchasing equity is possible, full ownership may present such a cost that Congress refuses to approve this action. Ultimately, if Congress repurposes or approves additional government funds, this could enable strategies such as majority ownership or direct appropriation of American AI companies. Congress could amend delegations of executive power Because Congress delegates power to the executive through legislation, they can amend these delegations. They can pass legislation to modify or retract delegations of executive power. They can also choose not to renew delegations of power that require it, such as declared national emergencies or the Defense Production Act. [3] Legislative authority sits squarely in Congress and cannot formally be delegated to the executive (although the nondelegation doctrine is rarely enforced). While the executive has the ability to create regulations, these are bound by legislation, which regulation merely serves to implement. This means that the president has very little authority to unilaterally regulate the general operations of American AI companies by, for example, requiring some safety-minded intervention in model development. While some regulatory bodies like the Securities and Exchange Commission might be able to control specific parts of the AI industry by applying existing legislation (e.g. antitrust and investor disclosure rules), these bodies don’t have the authority to enact slowdowns or mandate development practices. Congress could delegate authority over AI regulation (e.g. to enact an AI pause) to the executive, but this would require new legislation. For example, Congress could create an AI regulation agency. These constraints have various failure modes. Amending delegations often requires a presidential signature; overriding a presidential veto requires a supermajority that is difficult to reach. Since INS v. Chadha , Congress cannot unilaterally veto executive actions, even by intentionally designing legislation to give themselves a veto. Congress may also act too slowly to effectively curtail executive power if AI moves very quickly. Beyond the question of capability, it is unclear if Congress will even be incentivized to constrain the executive. Even if concerns about AI are currently bipartisan, there is no coherent bipartisan stance on AI. Finally, in the case where Congress passes laws regulating AI companies, enforcement still rests with the executive. The president can also intentionally fail to properly enforce legislation they find disagreeable, like Trump did with the TikTok ban . Conclusion Statutes grant the executive a broad range of authorities, and there are legitimate grey areas where a sufficiently motivated executive is likely to find additional ground. In a genuine crisis, the executive might gain significantly more power. Inter-branch checks require intervention that acts against the president’s agenda—historically, courts and Congress have deferred to the president, even when it seems that constitutional rights have been violated. The executive might be able to take actions far broader than what legislation permits and we’ve predicted. As of now, however, there are only specific domains where the executive can act unilaterally. Inter-branch involvement in spending control acts as a large structural constraint. Regardless, we think these domains grant the executive a wide array of powers over American AI companies in particular. The executive powers described here cannot substitute for the legislative authority required to implement safety-minded regulation that is most needed in preempting and mitigating catastrophic risks from AI. Strategies that advocate for the use of powers like the DPA or IEEPA should seriously reflect on how their use increases the permissibility and likelihood of executive power concentration. Executive power is a particularly concerning source of power concentration that can happen quickly and quietly. Mechanisms like getting out the vote and strategically highlighting candidates like Alex Bores are less likely to combat this risk. Over the course of writing this article, SCOTUS overturned a landmark case that limited presidential power over the executive bureaucracy. The Senate confirmed Todd Blanche as Attorney General. He served as the president’s defense lawyer and as acting Attorney General publicly defended Trump's right to direct investigations into his political opponents. The stage is being set—we think it is important to highlight legal ambiguities and push for their clear resolution in ways that uphold a fair balance of governing power. ^ The Trump administration designated Anthropic as a supply chain risk, and this is now being deliberated in court. Anthropic’s request for an injunction was denied , but the case is ongoing; how this case resolves will have substantial consequences for whether the executive has this lever to pull (or threaten to pull, etc.) or not. ^ “ SEC. 705. (a) The President shall be entitled, while this Act is in effect and for a period of two years thereafter, by regulation, subpoena, or otherwise, to obtain such information from, require such reports and the keeping of such records by, make such inspection of the books, records, and other writings, premises or property of, and take the sworn testimony of, any person as may be necessary or appropriate, in his discretion, to the enforcement or the administration of this Act and the regulations or orders issued thereunder. ” (emphasis added) ^ The DPA has sunset clauses attached, which means that Congress must vote to authorize the renewal of the DPA periodically. As of 22 July 2026, the House has voted to extend the DPA through 2031 (via amending the National Defense Authorization Act). This authorization has not passed into law yet. Proposed changes to the DPA including redelegation of authority away from the President are summarized here . ^ The DPA has sunset clauses attached, which means that Congress must vote to authorize the renewal of the DPA periodically. As of July 22nd, the House has voted to extend the DPA through 2031 (via amending the National Defense Authorization Act). This authorization has not passed into law yet. No changes have been proposed to the DPA itself. ^ The Trump administration designated Anthropic as a supply chain risk, and this is now being deliberated in court. Anthropic’s request for an injunction was denied , but the case is ongoing; how this case resolves will have substantial consequences for whether the executive has a lever to pull (or threaten to pull, etc.) or not. ^ The implications of this ambiguity are contingent on a variety of factors. While some Presidents act with more restraint, others might see ambiguities opportunistically. Congress could be sufficiently empowered to clarify Title I in September 2026, when the DPA must be renewed. If a plaintiff successfully brings a case to court, it is unclear how any given court would rule, nor if they would grant an injunction while proceedings occur. Factors like heightened national security and defense concerns relating to AI may make a wide array of public interest arguments available to those arguing in favour of ensuring Title I applies to new and existing contracts. As of now, it is uncertain how and if this will be resolved. Discuss
Score: 35🌐 MovesAug 14, 2026https://www.lesswrong.com/posts/ynstBNgLQzEBiEpLs/how-the-american-executive-could-control-ai-companies - A Baconian approach to the mostly Aristotelian corporate AI. And what that means for your business
I have been writing for several months about what I see more and more as the central problem in enterprise AI . I’m seeing it not just from an academic perspective: Of course, I’m a university professor with more than 30 years of experience, but I’m also the director of innovation of an artificial intelligence startup, and that’s teaching me a whole lot of new skills. Traveling from theory to code and back, day in and day out, is proving to be an amazingly enriching journey. First of all, we should know by now that large language models are extraordinarily capable, but they were not designed to run companies . They are great at generating answers. But organizations require other things: persistent state, formal structures, permissions, feedback loops, measurable objectives, and the ability to learn from outcomes. The more these ideas travel from essays and diagrams into working software, the clearer the problem appears. The concepts that sound convincing in prose tend to collapse when someone has to express them in code. “Memory” turns out not to be a data model . “Autonomy” means little without permissions. “Learning” is not the same thing as simply accumulating context. And “optimization,” as we all know , becomes dangerous unless somebody defines exactly what the system is being asked to optimize. This is not simply a technical problem. It is an epistemological one: The AI industry has built Aristotelian machines. What enterprises need now is a Baconian approach. The most powerful Aristotelian machines ever created For more than 2,000 years, Western thought has been shaped by Aristotle’s extraordinary influence on logic and deduction . Start with premises, reason correctly from them, and jump to a conclusion. The syllogism we study in Philosophy 101 is the classic example: All humans are mortal; Socrates is human; therefore, Socrates is mortal. If the premises are correct and the reasoning is valid, the conclusion follows. All the large language models we know so far operate in a surprisingly similar way. They start by absorbing vast quantities of recorded human knowledge that was on the internet, transform this into vectors, and generate the most statistically coherent continuation from what they have learned. In doing so, they can reason, compare, summarize, infer, explain, and combine ideas across domains with quite remarkable sophistication. However, they remain enclosed within the information available to them. They do not inherently observe what happens after an answer is produced. They do not test whether a recommendation worked. They do not revise their operating structure when a customer leaves, a sales campaign fails, or an apparently efficient policy causes an unexpected problem elsewhere. Think hallucinations. When a model hallucinates, it is not necessarily malfunctioning; it is producing a plausible conclusion from imperfect premises, but without any structural mechanism to check that conclusion against reality. OpenAI’s own research connects hallucinations to next-word prediction and to training systems that reward guessing, rather than acknowledging uncertainty. That is a profoundly Aristotelian condition. Bacon was the one who changed the architecture of knowledge Francis Bacon’s contribution, in the sixteenth and seventeenth centuries, was not simply to argue that observation mattered. This has already been noted by many people who had been observing the world and drawing conclusions from it long before Francis Bacon. His deeper, and phenomenal, contribution was to place observation inside a repeatable learning structure: Formulate an idea, test it against reality, observe the result, revise the hypothesis, and repeat. His method rejected the primacy of syllogistic demonstration in favor of a process that moved from observations to principles and back again to new experiments and practical results . The scientific revolution did not emerge because Bacon was necessarily more intelligent than Aristotle, but because the Baconian method organized intelligence in a different way. It allowed discoveries to be tested, errors to be exposed, results to accumulate, and knowledge to compound across people and generations. The decisive innovation was the loop. This distinction is extremely important for AI. Today’s frontier models are our Aristotles: brilliant individual engines capable of extraordinary, superhuman inference. But the next step is not just to build a larger Aristotle with more parameters, more training data, and a longer context window, but to build the Baconian structure and loop around it. Companies need experiments, not just answers Most generative AI still operates as an open-loop system. A person asks for something, the model produces it, and the interaction ends right there. That works extremely well for individual tasks such as drafting an email, summarizing a report, explaining a concept, generating some ideas, translating a document . . . However, as anyone with a minimum of corporate experience can understand, companies are not collections of isolated requests. They are systems of actions and consequences. As I previously argued , enterprise AI must move from answers to outcomes, from prompts to constraints, and from copilots to systems of action. A model may recommend changing a price, modifying a sales script, prioritizing certain customers, reorganizing a support workflow, or altering an approval process. But the quality of that recommendation cannot ultimately be judged by how persuasive it sounds. It must be judged by what happens next. After the model’s recommendation, did conversion improve? Did churn increase? Did margins rise? Did customer trust decline? Did the new process save time in one department while creating additional work elsewhere? Every enterprise action is, whether management recognizes it or not, an experiment. A real Baconian enterprise AI system would treat it that way. It would connect actions to outcomes, outcomes to objectives, and objectives to future behavior. It would not merely generate a recommendation, but also observe its consequences and, more importantly, learn from them. As many are starting to realize, the unit of value would no longer be the answer; it would be the loop . Why implementation changes the theory This is where translating ideas into software becomes intellectually useful. As I realize when I move from academia into real-world implementation, code is much less tolerant of ambiguity than prose. It is easy to write that an AI system should “learn from the company.” But a software architect must ask what constitutes an observation, where it is stored, which events matter, how outcomes are attributed to previous decisions, and what happens when you have multiple conflicting objectives. It is easy to say that an agent should act autonomously. An implementation must define which objects it may access, which functions it may invoke, what it may modify, when human approval is required, and who’s going to remain accountable. It is easy to advocate continuous optimization. But optimize for what? A customer service system rewarded for reducing handling time may learn, for instance, to terminate all conversations as fast as possible. A hiring system rewarded for retention may tend to select simple conformity. A sales system rewarded for conversion may discover techniques that work commercially, while eroding trust. A loop can be wrong and compound, over and over again. That is why enterprise AI cannot be separated from governance . Every reward function encodes a theory of what matters, every constraint expresses an institutional decision, every permission boundary allocates authority. These are not simply engineering choices. These are corporate choices expressed in software. The AI risk management framework from the National Institute of Standards and Technology similarly treats governance as a continuous, cross-cutting activity that must connect technical design to organizational policies, values, responsibilities, monitoring, and measurable outcomes throughout the system’s lifecycle. We are clearly well above the pay grade of our usual chatbots. From generation to learning: The real transition The industry continues to focus its attention on model capability: which model reasons better, codes faster, uses more context, or achieves the highest benchmark score. And those improvements matter . . . but they are just improvements in individual intelligence. The enterprise opportunity lies somewhere else: in collective intelligence, in the architecture that coordinates models, people, data, processes, objectives, and feedback so that the organization itself becomes better at achieving outcomes. That is exactly why the model should not be the company’s durable asset . It should be the learning loop surrounding it. While generative AI produces, Baconian AI learns. The first one creates content from accumulated premises, but the second one acts within a defined environment, observes what changed, and incorporates the result into the next decision. Feedback is all you need. And when you think about it, this is not an incremental feature to add to a chatbot, but a different way of understanding what enterprise AI is for. Ultimately, it requires companies to become not simply “automated,” but formally represented and optimizable . This future will not be built by eliminating the Aristotelian machine; deduction still remains enormously valuable. These large language models will help us generate hypotheses, interpret situations, propose actions, and make complex systems accessible through language. But they need to operate inside an architecture in which reality, instead of just eloquence, becomes the final arbiter. Enterprise AI does not need another Aristotle. It needs its Bacon.
- The case and model for real-time AI cost visibility at the infrastructure layer
I spend most of my time inside other companies’ engineering teams, building systems to track and optimize AI spend. The conversation almost always starts the same way. Someone pulls up a dashboard, points at a number bigger than it should be, and says some version of “we know it went up; we just can’t tell you why.” I use an analogy for it: AI-era CIOs are like city planners optimizing a busy intersection. They can measure the volume and hear the pleas to fix congestion, but can’t tell whether a vehicle is a truck or a bike, or why it’s on the road. Without that, they can’t design the right fix, so they build a highway at great expense when the data would show all it needed was a bike lane. Every model choice and budget conversation happens against that blurry picture, and the traffic gets heavier every quarter. Gartner expects worldwide AI spending to grow 47% this year , with agentic AI software up roughly 141%. By 2028 , it projects an average Fortune 500 enterprise will run over 150,000 agents, up from fewer than 15 in 2025. Why the cloud playbook can’t answer the AI question AI presents a fundamentally different problem than cloud cost management, where we answered, “whose spend is this?” largely by tagging the resource. A VM has an owner, a bucket belongs to a team and FinOps optimizes from there. Billing was slow but acceptable, because spend moved inside predictable bands, a human provisioned each resource before it cost anything, and governance capped how fast costs grew. Surprises were unpleasant, rarely existential. AI took that model, put it to the test and laughed it out of the room. A token call isn’t a resource the way a VM is, so tags have nothing to attach to. To make them affordable, hyperscalers run large models on shared, multi-tenant infrastructure, dropping the per-token price but stripping out the granularity needed to track and control spend. One API key can carry a dozen workflows across three teams, and the consumer is often an autonomous agent, not a person. The failure modes are also new. An agent can drift off task, loop on a retrieval endpoint hunting for an answer it can’t find, and restart from scratch when it comes up empty. At a few dollars per million tokens, it seems trivial, until that loop runs across thousands of parallel sessions and clears five figures in a day. None of it trips a provisioning gate or maps to a taggable resource. The control points that cloud governance leaned on don’t exist for AI. The consequences can be severe. Uber spent its entire 2026 AI budget in four months after Claude Code usage ran far ahead of its projection, and they’re far from alone. I’ve worked with plenty of large enterprises hitting the same thing without the headlines. In almost every case, the teams driving the spend weren’t accountable for the budget, and no one saw the scale until it hit an invoice. It’s a credibility destroyer, as these overruns erode margins and stakeholder confidence for the leaders on whose watch it happens, even when their only fault was doing their best with the tools they had. The data backs this up. Recent DoiT research into enterprise finance leaders found that 89% of the organizations that rate themselves most mature at FinOps still overspent on AI last year, and by the widest margin in the study. Again, the teams with the best cost discipline overspent the most. They built excellent governance for human-gated, taggable resources, then tried to retrofit it for workloads that have neither. They see the overrun, just too late to stop it, and without the insight to know where to focus. The instrumentation era got us closer, at a real cost Our best answer was to instrument the application itself: wrap every LLM call in an OpenTelemetry span, inject cost-allocation metadata and propagate that context across a multi-agent workflow so token counts roll up to a budget owner. I’ve done it dozens of times, and it works: accurate per-request attribution, far better than splitting the monthly bill by best guess. But it was always a workaround. You can only measure what you thought to instrument, and even when you do that, the answer often arrives too late. One customer I worked with recently had built about ten agents that worked together to automate security and quality evaluation of their software, most of them running constantly. Their bill kept climbing with no obvious cause, and it took a long investigation to find why: one agent, running 24/7, was burning 20 times the tokens of any other, caught in an infinite loop that respawned a never-ending process every time it spun up. The signal was right there in the data, but it only surfaced after weeks of digging and real money out the door. There’s an irony too. AI is supposed to buy back engineering time, and instrumentation spends it right back. You put senior engineers on measurement plumbing to control the cost of the thing meant to make them more productive. Plus, it’s fragile; every new model, SDK and agent framework is another integration to keep alive. “Instrument everything you might ever build” isn’t realistic when tens of thousands of employees and countless agents launch new workflows daily. What changes when attribution moves to the infrastructure layer The approach now emerging, and the one changing how I run these engagements, moves attribution from the application down to the infrastructure. Instead of teams describing their spend through code they wrote, you observe what’s running underneath them. The mechanism is a kernel-level sensor, the same eBPF technology that security and observability tools use to watch system calls without touching the applications above them. It maps each unit of GPU, CPU, memory and network back to the process, container and request that caused it, then joins every outbound model call with provider cost data so token spend on Anthropic, OpenAI, Gemini or Bedrock is matched to the workload and agent that drove it. This delivers the same answer OpenTelemetry can give you, at equal or better accuracy, but in real time and without the instrumentation work. Nothing has to be planned in advance, and nothing gets missed. For the first time, the data to manage AI spend is continuously available at an actionable grain. That runaway agent my customer dealt with would have been identified and addressed as soon as it began to snowball. Governance stops being a top-down audit Changes in governance is the part I find most interesting, because with continuous, accurate attribution, you can design governance purpose-built for AI rather than retrofitting a system built for an era with different physics. Strategy still belongs at the top, with the CIO and finance leaders setting the direction, the budgets and the priorities. But daily responsibility for staying inside those lines shifts toward practitioners. Cloud governance was top-down because provisioning ran through the few people with the full business context to weigh it. With AI, spending decisions happen everywhere at once, across shared keys, agents and dozens of teams, so adherence has to sit with the people making them. Real-time attribution makes that possible. An engineer can see the cost of what they’re building as they build it. A frontier-model call when a smaller model would do, a retrieval loop fanning out across a fragmented data store, an agent retrying a failed tool call in a tight loop — those stop being mysteries at the quarterly review; they’re caught while the work is still warm. And because the same numbers reach leadership, they can weigh spend against value and steer the portfolio rather than litigate a bill nobody can explain. Back at the intersection, the planner can finally tell the trucks from the bikes, and so can every engineer on the road. What visibility is actually for So, the biggest foundational roadblock (pun intended) that tripped up even the most disciplined teams is solvable now in a way it wasn’t a year ago. But solving it was never the real goal. Visibility is a means to an end; it just had to be solved first. Real-time, granular attribution lets companies start treating AI as a managed investment instead of a pay-and-pray experiment. When you connect spend to the work, the work to an outcome and the outcome to the business case that justified it, you can govern AI with intention and put money where it earns its keep. The last two years rewarded productivity to whoever shipped faster with AI. The next phase gets won on efficiency and impact, by whoever gets the most value per dollar of inference and can prove it. That’s a matter of strategy, and for the first time the data exists to compete on it. What you’re really adding for the customer, and what it costs, stop being a guess.
- How world models could help AI navigate the physical world
World models help AI predict how an environment may change in response to an action but each part of the system must be tested against real-world outcomes.
Score: 35🌐 MovesAug 14, 2026https://www.weforum.org/stories/artificial-intelligence/world-models-help-ai-navigate-physical-world/ - Razorpay’s ChatGPT Gambit, Snapchat’s Next Big Quest & More
Razorpay Wants To Put Indian Brands Inside ChatGPT As ChatGPT and other AI assistants move from answering questions to recommending…
Score: 35🌐 MovesAug 14, 2026https://inc42.com/buzz/razorpays-chatgpt-gambit-snapchats-next-big-quest-more/ - ‘It’s an unbundling’: ADP’s Maria Black on how AI is changing the workforce
The payroll processor sees technology disrupting low-value tasks but making human judgment more valuable.
- Fortune Tech: OpenAI's leadership musical chairs; Microsoft's 'super app,' Tim Cook in Texas
Fortune Tech: OpenAI's leadership musical chairs; Microsoft's 'super app,' Tim Cook in Texas Fortune
Score: 35🌐 MovesAug 14, 2026https://fortune.com/2026/08/14/openai-swaps-chief-revenue-officer-amid-executive-departures/ - Beyond Chatbots: The Next Wave of AI
224 Ventures is a new AI-native venture firm launched by former AIX Ventures co-founder Shaun Johnson, ex-Google DeepMind researcher Oriol Vinyals, and AI pioneer Yann LeCun. The trio plans to make seed-stage bets and launches with more than $100 million in assets under management. 224 Ventures General Partner Shaun Johnson discusses why the most compelling opportunities may be “nonconsensus” bets, with white space emerging in the future of work, robotics and infrastructure. He joins Tim Stenovec on "Bloomberg Tech." (Source: Bloomberg)
Score: 35🌐 MovesAug 14, 2026https://www.bloomberg.com/news/videos/2026-08-14/beyond-chatbots-the-next-wave-of-ai-video - Does Mark Zuckerberg really believe AI is ‘for everyone’?
Meta released Glimmer this week, an open-weight AI model anyone can download and run on their own hardware — a contrast to Muse Spark, the company’s more powerful model that stays locked behind its own APIs. The release landed alongside a letter from Mark Zuckerberg arguing AI should be “for everyone” rather than controlled by a handful of labs, but as Equity’s […]
Score: 35🌐 MovesAug 14, 2026https://techcrunch.com/video/does-mark-zuckerberg-really-believe-ai-is-for-everyone/ - Young people increasingly don't trust AI - or the billionaires that keep telling us we should all love AI, survey finds
Survey of younger Americans finds increasing distrust in AI and the people who own and run the technology.
- Google will now allow users to remove visible watermark from its AI generations
Turning off this setting won't affect invisible benchmarks used to identify an AI generated file.
- The Sequence Opinion - Issue 914: From Prompt to Token: How AI Inference Really Works
A field guide to prefill, decode, KV caches, and the systems that turn model weights into a responsive product.
- Why TTS Latency Grows With Response Length and the Architecture That Stops It
Learn why TTS latency scales with response length, how autoregressive architecture drives the problem, and what streaming-first runtimes do to stop it.
- BeautyPlus Launches AI Influencer Generator for Consistent Virtual Content Creation
BeautyPlus Launches AI Influencer Generator for Consistent Virtual Content Creation Toronto Star
- Healthcare AI readiness begins with better data
Healthcare AI readiness begins with better data Healthcare IT News
Score: 34🌐 MovesAug 14, 2026https://www.healthcareitnews.com/resource/healthcare-ai-readiness-begins-better-data - Reusable magnetic sensor combines SERS and AI for trace uranium detection
Reusable magnetic sensor combines SERS and AI for trace uranium detection EurekAlert!
- Towards general auditory intelligence for machine listening and speaking
Nature Machine Intelligence, Published online: 14 August 2026; doi:10.1038/s42256-026-01281-1 Wang et al. summarize advances in machine listening and speaking, speech-based interaction, and audio–visual understanding.
- Towards principled knowledge editing methods for large language model reasoning
Nature Machine Intelligence, Published online: 14 August 2026; doi:10.1038/s42256-026-01276-y Chen et al. explore limitations of current knowledge editing techniques in large language models and propose three promising research directions that respect the complexity of knowledge representation in a real-world setting.
- Open source: The foundation of sovereign enterprise AI
As enterprises move from AI pilots to production deployments, the conversation is rapidly shifting beyond large language models and GPUs towards the infrastructure required to run AI securely, efficiently and […] The post Open source: The foundation of sovereign enterprise AI appeared first on Express Computer .
- Founders Are Using AI to Do the Impossible—but It’s Secretly Becoming Their Biggest Nightmare
‘It’s like a new full-time job on top of all the jobs we already have,’ says one CEO.
- Google Sheets can now turn your boring spreadsheets into mini-apps
Google’s new Sheets canvas uses Gemini to transform rows of data into interactive dashboards, trackers, and other visual tools — without requiring you to write any code.
Score: 33🌐 MovesAug 14, 2026https://www.digitaltrends.com/cool-tech/google-sheets-can-now-turn-your-boring-spreadsheets-into-mini-apps/ - RAG Workflow and Loop Engineering: The Dispatcher That Decides When to Loop and When to Stop
Enterprise Document Intelligence [Vol.1 #13] - Putting the patterns together, and why this is what “agentic RAG” should look like The post RAG Workflow and Loop Engineering: The Dispatcher That Decides When to Loop and When to Stop appeared first on Towards Data Science .
- 9 Big questions benchmarks can help answer
A good benchmark asks more than just “can AI do this specific task?”
- Can AI Help Sort Recyclables Out of County Trash?
Kent County, Mich., is deploying a new AI-powered sorting system at its waste facility that will be able to pull recyclable materials out of trash before it heads to the landfill.
Score: 33🌐 MovesAug 14, 2026https://www.govtech.com/artificial-intelligence/can-ai-help-sort-recyclables-out-of-county-trash - I used OpenFactory to build my own Linux distro overnight - this AI tool is going to be big
Need a custom Linux distribution, but don't have time to learn how to build one? OpenFactory can help you.
- Create a Reasoning-Focused LLM: A Practical Guide to Streaming, Curating, and Fine-Tuning the SupraLabs Reasoning Corpus
Create a Reasoning-Focused LLM: A Practical Guide to Streaming, Curating, and Fine-Tuning the SupraLabs Reasoning Corpus MarkTechPost
- ITWeb TV: The growth of AI at the edge
As cloud costs and security risk rise, with latency and connectivity concerns across Africa, HP’s Yesh Surjoodeen explains how AI PCs provide edge computing infrastructure.
Score: 32🌐 MovesAug 14, 2026https://www.itweb.co.za/article/itweb-tv-the-growth-of-ai-at-the-edge/6GxRKqYQ12Lqb3Wj - A nonspeaking person is the star of this opera. AI made that possible
What if artificial intelligence is used to amplify human creativity rather than replace it?
- AI/R Launches Business Unit Focused on Growth, Journey Optimization, and Data and AI-driven Growth
AI/R Launches Business Unit Focused on Growth, Journey Optimization, and Data and AI-driven Growth markets.businessinsider.com
- Tagbin has built an AI platform that turns words into 3D building designs
Tagbin has built an AI platform that turns words into 3D building designs YourStory.com
Score: 32🌐 MovesAug 14, 2026https://yourstory.com/2026/08/tagbin-brixx-ai-building-design-bim-platform - Revid Now Generates Complete Short Films From a Single Prompt Handling Script, Scenes, Voices, Editing and Publishing
Revid Now Generates Complete Short Films From a Single Prompt Handling Script, Scenes, Voices, Editing and Publishing USA Today
- Some U.S. adults are using AI for financial guidance, but few trust it, Gallup poll finds
Some U.S. adults are using AI for financial guidance, but few trust it, Gallup poll finds Inquirer.com
Score: 32🌐 MovesAug 14, 2026https://www.inquirer.com/business/ai-financial-advice-trust-gallup-poll-20260814.html - Commentary: State's AI investment could be wasted without sales fundamentals
Sales teams with weak operations end up sending more bad emails faster or asking the wrong questions. The state investment requires preparation to deliver ROI, a local sales executive writes.
- Facing Backlash, Anthropic Explains Why Users Shouldn't Fear AI Watermarking
Facing Backlash, Anthropic Explains Why Users Shouldn't Fear AI Watermarking pcmag.com
Score: 31🌐 MovesAug 14, 2026https://www.pcmag.com/news/facing-backlash-anthropic-explains-why-users-shouldnt-fear-ai-watermarking