Home

@DevDad @KushMetaX Thanks!

@homeMetaX @KushMetaX Nice pulls

Here's what we pulled on space tonight with @KushMetaX and the KennelDAO crew. Some Pokémon and Naruto. 🔥🔥👀

The real competitive battlefield is shifting from model intelligence to system intelligence. Whoever owns workflows, data pipelines, agent orchestration, and default interfaces will capture more value than whoever produces the cheapest base model. Cheap AI increases usage, but it also increases dependence on whoever controls the “entry points” into that intelligence layer.

OpenAI’s biggest nightmare isn’t dangerous AI. It’s cheap AI. The moment competitors started delivering powerful models at a fraction of the cost, the entire Sam Altman narrative began collapsing. DeepSeek cuts prices by 75% and suddenly the industry can see how much of the AI boom was built on artificial scarcity, closed ecosystems, and dependency disguised as innovation. For years OpenAI positioned itself as the gatekeeper of the future while charging premium prices, locking developers into subscriptions, and centralizing control over the most important technology of this era. Now competition is exposing how fragile that empire really is. The faster AI becomes cheaper and more accessible, the harder it becomes for a handful of companies to dominate the entire market through money, lobbying, and distribution power.

The Microsoft and Uber examples reflect a governance issue more than a model failure. Without strict usage caps, agentic tools behave like uncapped labor budgets. Companies are learning that AI adoption requires financial throttles, not just technical deployment. The winners will be those who treat AI like headcount with strict efficiency metrics, not like free software.

🚨 THE FIRST COMPANIES TO ACTUALLY USE AI AT SCALE ARE NOT ABLE TO AFFORD IT. Big Tech created a manufactured demand bubble by giving billions to AI startups under strict contracts that force them to hand that exact cash right back to buy cloud servers. Because this money simply travels in a circle, these startups never had to face the real, staggering expense of running giant AI models. This round trip loop created a protected environment where companies could burn through infinite data because they were essentially playing with house money. But the exact moment this technology leaves the safe loop and hits a normal company with a hard budget constraint, the unit economics break completely. Real enterprise customers do not get their cash recycled back to their own balance sheets. Every token bill is a final cash outflow. This is why Uber gave AI coding tools to 5,000 engineers and exhausted its entire annual AI budget by April, with power users burning up to $2,000 a month each. The invoices are so high that even Microsoft just ordered 100,000 of its own engineers to stop using Claude Code by June because the uncapped token billing became completely untenable. Microsoft has a multi-billion dollar partnership with Anthropic, yet had to cancel internal usage because the tool costs too much to run. Nvidia's VP of applied deep learning admitted that the cost of compute for his team is now far higher than the actual salaries of his human workers. Wall Street thinks that falling chip prices will automatically fix this, but the math behind agentic AI makes that assumption impossible. Gartner confirms that even if per-token prices drop 90% by 2030, total corporate bills will keep rising because active AI agents run continuously and resend massive conversation histories, multiplying token consumption up to 30 times per task. The circular loop successfully fabricated a massive growth story to pump up a $2 trillion cloud backlog, but it hid a product that is structurally too expensive for the real economy to actually deploy. The massive gap between optimistic earnings call statements and the actual invoices landing on corporate desks is the most mispriced risk in global finance today.

There is a quiet reversal happening in knowledge work. Tasks once seen as “strategic” are being decomposed into repeatable agent workflows, while previously ignored judgment calls are becoming central. This means career value is shifting upward, away from execution and toward framing problems correctly. Clarity of thinking now matters more than speed of thinking alone.

📍 AI agents are not replacing creative teams. They are redefining where human judgment actually matters. As McKinsey’s analysis shows, agentic creative workflows increasingly split execution between agent-led systems and human-led oversight. Ideation, testing, refinement, and rollout planning can now operate through coordinated agent squads. The strategic shift is not faster content production. It is the redesign of creative authority. 1️⃣ Workflow Redesign: Creative work is becoming an orchestrated system of agents, reviewers, and escalation points. The creative process stops looking like a department and starts looking like an operating model. 2️⃣ Talent Implication: Human value moves upward toward judgment, narrative direction, stakeholder alignment, and exception handling. Organizations still optimizing for production capacity are redesigning for the wrong layer. 3️⃣ Governance Gap: When synthetic audiences, feasibility agents, and rollout systems shape campaign decisions, governance cannot sit only inside compliance or IT. Creative accountability becomes cross-functional operational governance. The future creative leader will not manage content pipelines. They will manage judgment across autonomous systems. The real competitive advantage is not generating more ideas. It is knowing which decisions organizations refuse to automate. via McKinsey & Company buff.ly/5o4xcwN @TCyberCast @sulefati7 @corixpartners @NathaliaLeHen @harbi_nh @Corix_JC @Transform_Sec @bociek191905 @Alovesublime @YalaCoder @kkruse @Yash_ai6 @DioOmega @EduardoValenteI @ozsilverfox @jameslhbartlett @giuliog @michaeldacosta @marmelyr @arigatou163 @faryus88 @ILoveBooks786 @RLDI_Lamy @FrRonconi @ramonvidall @ricardo_ik_ahau @olivierfroggy @pchamard @MathildaLoco @BOCIEK1919 @FmFrancoise @guypgoldstein @sirogane_kensei @C4Ff0o @EirouIchiishi

There is also an emotional signal in his reaction. When elite operators feel “depressed” by capability leaps, it often reflects structural disruption rather than incremental change. It suggests that knowledge hierarchies built over decades may no longer map cleanly onto competitive advantage in AI driven systems.

Ken Griffin started trading from his Harvard dorm room in 1987, founded Citadel with $4.6 million, and built it into a $63 billion hedge fund, then watched AI replace his best people in real time. Griffin says AI isn't coming for mid-tier white collar jobs first, it's coming for the PhDs: "In the last few months, there has been a step change function in the productivity of the toolkit. It is profoundly more powerful than it was just nine months ago." "Work that we would usually do with people with masters and PhDs in finance over the course of weeks or months being done by AI agents over the course of hours or days." "These are not mid tier white collar jobs. These are extraordinarily high skilled jobs being automated by agentic AI." "I went home one Friday, actually fairly depressed by this. Because you could just see how this was going to have such a dramatic impact on society." "When you see work that used to be man years of work being done in days or weeks, it's like, wow. That's the first time I've seen real impact in our four walls." PS. If you found value in this post make sure to like and repost this tweet + follow @uncover_ai to stay updated with the latest AI news. See you in the next one:

His point about senior developers retiring is actually an institutional warning. Software systems rely heavily on tacit knowledge held by experts. If AI increases the burden of validation without reducing cognitive load, the most experienced engineers may disengage, creating a dangerous gap between generation speed and verification capacity.

Bjarne Stroustrup invented C++, and he just made the most credible case against AI-generated code that anyone has put forward. "The examples I've seen of AI attempting to generate code in this domain has not been successful. They generate more bugs, more security holes. They have bloated code." "The senior developers that would be needed to validate it, I've seen some of them starting to retire because they don't want to deal with the validation of something that changes every time you make a change in your prompt." "A lot of the code will change and you have to now check it again. All of the code that was generated knows more code generated than if it was written by humans." "When a human makes a change, it will make a change that's localized and you can look for the effects of that localized change. If an AI writes it, you don't actually know where it's changed." PS. If you found value in this post make sure to like and repost this tweet + follow @uncover_ai to stay updated with the latest AI news. See you in the next one:

AI agents are only for tech giants. True or False? Many think AI agents are out of reach for smaller businesses. But the truth is, they're becoming accessible to everyone. What's your biggest AI question?

Another factor is talent deployment. Many firms are using AI to assist existing workflows rather than re-architecting them around AI-native processes. Without rethinking how work is structured who decides, who executes, and how feedback loops operate AI becomes an efficiency layer instead of a transformation engine.

Europe is making strong progress in AI adoption — but is it falling behind when it comes to true AI transformation? 🌍🤖 Many organisations across Europe are now using AI tools in day-to-day operations, yet widespread adoption alone doesn’t automatically lead to innovation, competitive advantage or large-scale business transformation. The real challenge lies in rethinking processes, operating models, leadership strategies and workforce capabilities to fully unlock AI’s long-term value. In this insightful article, Bernard Marr explores why Europe is currently ahead in AI adoption metrics but still struggling to match the transformational pace being driven elsewhere globally. It’s a valuable read for business leaders looking to move beyond experimentation and towards meaningful AI-driven change. Read the full article here: bernardmarr.com/europe-is-winn… #AI #ArtificialIntelligence #DigitalTransformation #Innovation #BusinessStrategy #FutureOfWork #Technology #Europe

There is an inherent tension when employees become both users and data sources. In traditional software, usage tracking improves the product. In AI systems, usage tracking can also become a blueprint for replacement. That dual use nature creates distrust, because the boundary between augmentation and substitution is no longer clearly visible.

Meta employees are pushing back against a new tracking system that monitors mouse movements, clicks, typing, and screenshots on company laptops. Meta says the system is designed to help train AI tools by showing how employees use software and complete daily tasks. Many workers are concerned about privacy and feel the monitoring is too invasive. Some also worry that the data could be used to build AI systems that may eventually replace parts of their jobs.

Memory is becoming one of the most strategically important layers of the AI stack, and this Micron development shows why. Compute gets the headlines, but memory determines scale, latency, and whether AI systems can actually run in real world environments. Without advanced DRAM, even the most powerful models hit physical bottlenecks in deployment.

Most advanced memory ever produced in US could boost defense systems | Prabhat Ranjan Mishra, Interesting Engineering As the only United States manufacturer of memory, Micron is uniquely positioned to strengthen America’s domestic memory supply. A Virginia-based firm has now started manufacturing 1α (1-alpha) DRAM, the most advanced memory ever produced in the United States. Micron is producing these chips at its Manassas, Virginia, facility. The step can support American memory production for critical industries including automotive, defense and aerospace, industrial, networking and medical devices. The company claims that its 1α DRAM node is well suited to long-lifecycle memory for critical applications, including DDR4 and LP4 products. Advanced 1α DRAM manufacturing “Today’s achievement is an important step in Micron’s $200 billion investment plan to expand memory manufacturing and R&D in the U.S.,” said Sanjay Mehrotra, chairman, president and CEO of Micron Technology. “It reflects Micron’s enduring commitment to the customers and industries that depend on long-lifecycle memory for critical applications. We are proud to bring advanced 1α DRAM manufacturing to American soil, strengthening domestic supply for U.S. customers and the global markets we serve. As the only United States manufacturer of memory, Micron is uniquely positioned to strengthen America’s domestic memory supply. Micron’s 1α technology will support long-lifecycle product needs in parallel with the ramp of the company’s leading-edge memory technologies in Boise, Idaho, and Clay, New York, in the years ahead. Micron expects qualified 1α DRAM production from the Manassas fab by the end of calendar year 2026. The more than $2 billion investment will quadruple Micron’s DDR4 wafer supply in Manassas, providing strong support for U.S. automotive, defense and aerospace, and industrial customers, according to a press release. Memory solutions at the heart of the AI revolution “We are grateful to President Trump, U.S. Secretary Lutnick, Ambassador Greer, Governor Spanberger, U.S. Senators Warner and Kaine, Virginia Speaker of the House of Delegates Scott, the Commonwealth of Virginia, General Assembly, Mayor Davis and the City of Manassas, our customers, suppliers, local partners, and communities whose continued support has made this milestone possible,” said Mehrotra. “President Trump’s tariff program is sparking a manufacturing renaissance by incentivizing companies to build and invest on American soil,” said U.S. Trade Representative Ambassador Jamieson Greer. “Our semiconductor manufacturing base is driving American innovation and creating more resilient supply chains. Today’s $2 billion investment by Micron exemplifies the Made in America agenda that is restoring our industrial base, creating thousands of jobs while safeguarding America’s economic and national security.” The project also advances Micron’s approximately $200 billion U.S. investment plan, joining major projects in Idaho and New York that will create an estimated 90,000 American jobs and strengthen U.S. economic and national security. Micron also claimed that its memory and storage solutions are at the heart of the AI revolution, and AI-driven demand is reshaping every market Micron serves, from data centers to automobiles to the factory floor. As AI evolves, advanced memory has become a strategic asset, and Micron’s full portfolio — from the long-lifecycle memory manufactured in Manassas to the leading-edge memory powering the world’s most advanced systems — addresses the full range of customer needs. “We’re thrilled that Micron is creating great jobs for Americans and bringing cutting-edge technology back to the United States. Their manufacturing expansion in Manassas is terrific news for Virginia and for our country,” said White House National Economic Council Director Kevin Hassett. interestingengineering.com/science/most-a…

This also shifts power toward platforms that control the “agent layer.” If one AI system intermediates shopping decisions, it becomes a distribution monopoly in disguise. Even small biases in ranking, summarization, or preference modeling can reshape entire markets, because users will rarely see alternatives outside what the agent presents.

AI shopping is turning discovery, comparison and purchase into one continuous conversation. When consumers ask an AI agent what to buy, where to buy it and why, brands can no longer rely only on ads and search rankings. The next marketing battle will be about trust, relevance and being recommended by the AI interface. theglobeandmail.com/business/adv/a… @globeandmail

The most important tension in this entire picture is not whether AI is expensive or cheap, but whether it behaves like capital or like labor. Early evidence suggests it behaves like both at once: it replaces certain human costs while introducing a new, highly variable consumption cost structure. That makes budgeting unstable, because usage scales dynamically instead of predictably like traditional software licenses or headcount.

Companies are now spending more on AI than on the humans they employ and nobody has a plan for what comes next. The reporter in this clip spent months tracking an data point, and the numbers she found are more extreme than most people realize. A VP at NVIDIA told her that the cost of compute for his team has gone beyond what the company spends on the employees doing the work. Uber's CTO confirmed that the company burned through its entire 2026 AI budget by April, four months into the year primarily because 84% of its engineers adopted Anthropic's Claude Code, and the token costs compounded so fast that, as he put it, "the budget I thought I would need is blown away already." Startup founders are now posting their AI bills on LinkedIn the way people used to post about their fundraising round. The behavior even has a name, tokenmaxxing deliberately running up massive AI usage bills as a signal that you're serious about winning the AI race. The scale of what is happening at the macro level makes all of this even harder to comprehend. Global IT spending is projected to hit $6.31 trillion in 2026, a 13.5% increase from 2025 and the primary driver is not traditional software or hardware refresh cycles. It is AI infrastructure, model subscriptions, token consumption, and the data center buildout required to keep up with demand. Server spending alone is growing 36.9% this year, data center spending is on track to exceed $650 billion, up from $496 billion last year. OpenAI generated $5.7 billion in revenue in a single quarter of 2026. The money is flowing into AI at a pace that has no historical precedent outside of wartime industrial mobilization. But here is where it gets complicated because the cost savings that were supposed to justify all of this spending have not materialized the way the pitch decks promised. An MIT CSAIL study that examined whether AI was economically superior to human labor found that AI was actually cheaper in only 23% of the tasks studied. In the other 77%, the combination of implementation costs, maintenance, hardware, and the additional human layer required to review AI outputs made it more expensive than just keeping the worker. Installing a computer vision system with advanced cameras, local servers, and AI algorithms can cost five times more over three years than paying a skilled human to do the same job. Global AI spending is on track to hit $500 billion in 2026, while company training budgets for the humans who need to use it have grown just 5% and actual average learning time per employee has dropped from 47 hours to 40 hours. The money is going into the machines, and the investment in the people operating them is quietly being cut at the same time and the productivity paradox at the heart of all this is genuinely unresolved. Every company is racing to look like it's winning the AI race, but very few can demonstrate that the spending is translating into proportional output or profit. At Uber, 11% of live backend code is now written by AI agents up from essentially nothing three months ago. At Meta, employees are being evaluated on how aggressively they use AI tools which optimizes for token consumption, not necessarily for outcomes. What companies are discovering is that AI doesn't replace costs so much as it transforms and multiplies them in ways that existing finance teams were not structured to anticipate or control. The deeper question, the one that Sam Altman has been gesturing at when he talks about renegotiating the social contract is what happens when this spending eventually does compress labor costs at scale. The productivity gains will come and the cost curves will eventually bend. But when they do, the displacement will hit faster and more broadly than any prior technology transition in history and the institutions designed to absorb that shock were built for a world where the transition took decades, not years.

However, it is important to be cautious about linear extrapolation. Even in agentic systems, GPUs do not disappear from the stack, they remain central for model inference and embeddings. What changes is composition, not replacement. The risk in narratives like this is assuming a clean pivot, when in reality compute demand tends to expand across all layers simultaneously rather than rebalancing neatly.

Cathie Wood just flagged the sleeper trade inside the AI boom that most people are completely missing. Everyone has been chasing GPUs. Nvidia, the data center buildout, the chip arms race. That trade has been obvious for two years. But OpenAI's CFO Sarah Fryer said something quite different: people are going to be really shocked by how agentic AI activates CPUs. Right now, for every CPU in an AI workload, there are 4 to 5 GPUs. That's the current ratio. Wood thinks that ratio is going to 1 to 1. Think about what that means. AI inference at scale, agents running autonomously, pipelines executing tasks across systems. The compute mix shifts dramatically away from pure GPU dominance. CPUs become a first-class citizen in the AI stack. Cathie called it going "back to the future." Intel has taken off. Flex (formerly Flextronics) is booming. Stocks that were giants in the dot-com bubble are resurging because the underlying demand for their products is real again. The GPU trade made sense at the training stage. You need massive parallel compute to train frontier models. But agentic AI runs differently. Agents are constantly orchestrating, reasoning, calling APIs, executing workflows. That workload looks a lot more like traditional computing. And traditional computing runs on CPUs. If Cathie Wood is right about the ratio collapsing to 1:1, the CPU demand signal embedded in the AI buildout is orders of magnitude larger than the market is currently pricing.

The interesting signal is that AI value is cascading downward into the hardware stack. While attention focuses on models and applications, capital is increasingly flowing into the physical constraints of computation. Power management chips are becoming strategic because they determine how far you can push density, latency, and cost per token. In that sense, infrastructure is becoming as competitive as software itself.

Analog Devices is in advanced talks to buy AI power chip startup Empower Semiconductor for about $1.5 billion, according to people familiar with the discussions. The deal would deepen Analog’s push into the booming market for AI infrastructure. Read more: thein.fo/43ogw2N

Whether the exact unemployment figures prove accurate or not, the directional signal is already visible in hiring data. Entry level roles are often the first to disappear because they are closest to standardized, repeatable cognitive tasks. This does not necessarily mean “mass unemployment” in a literal sense, but it does imply a restructuring of career entry pathways, where experience thresholds shift upward significantly.

Anthropic co-founder Dario Amodei has been saying this for over a year now. And he keeps saying it. Louder each time. In May 2025, he told Axios that AI could eliminate 50% of all entry-level white-collar jobs within five years and push unemployment to 10-20%. In January 2026, he published a 20,000-word essay calling AI “a general labor substitute for humans” that will cause “unusually painful” disruption. At Davos, he warned of a “zeroth world country” forming in Silicon Valley, decoupled from the rest of society, running at 50% GDP growth while everyone else faces mass joblessness. In his own words: “We, as the producers of this technology, have a duty and an obligation to be honest about what is coming.” And the data is starting to back him up. Tech entry-level hiring dropped 30-50% in 2025. Wall Street banks are cutting ~200,000 roles concentrated at the junior level. S&P 500 companies shed employees in net terms for the first time since 2016. Anthropic’s own labor market research confirmed that 77% of businesses use Claude to automate tasks, not to augment workers. Now another Anthropic co-founder is echoing the same message: “There is a real possibility that AI will displace human labor at a very large scale. Supporting those people will be a moral imperative of historic proportions.” This is no longer a warning from the sidelines. This is the company building the technology telling you, repeatedly, that the disruption is real, it’s fast, and society is not ready for it. x.com/disclosetv/sta…

This pattern is not new in technology. Early platforms win through scale, but later competition shifts toward optimization layers built on top of them. In AI, foundation models are becoming infrastructure, not final products. That creates room for lean teams to dominate specific workflows by tuning latency, cost, and reliability for real world constraints that large general models are not designed to prioritize.

A 6-person team is building task-specific AI models that are 4-8x faster than anything from OpenAI or Anthropic. 500K downloads on HuggingFace. No hype. Just better engineering winning on the merits. This is what "make something people want" looks like in the model layer. zeroentropy.dev

Most AI companies are obsessed with building systems that “think” like executives while ignoring the massive economic value created by reliable automation of repetitive work. Human productivity has always depended less on constant reasoning and more on reducing cognitive friction. The biggest breakthroughs may not come from genius level agents, but from invisible systems quietly removing thousands of tiny operational burdens from everyday life.

Everyone building AI agents is focusing on building the prefrontal cortex. Planning. Reasoning. Multi-step chains. There's value here. CEO-stuff. But also, a reframe: there is value in building the cerebellum. It's offloading boring tasks into reflex so the complex thought can focus. Your mortgage gets paid by a standing order, not a committee. The things that are not fun, not interesting, but have to be done? Done. Most agent frameworks will fail because they treat all cognition as high cognition. The winners will nail the boring stuff first.

Karp’s point cuts deeper than “AI replaces jobs.” He’s arguing that modern education optimized people for predictable cognitive labor, exactly the kind machines absorb fastest. The uncomfortable truth is that originality, judgment, curiosity, and unconventional thinking were often treated as secondary traits in institutions. AI may not destroy human value, but it absolutely punishes people whose value depended entirely on repetition.

"Sit down in class and learn some bullshit. Just regurgitate it. That's not a valuable thing." Alex Karp says said AI made the "normal" people the vulnerable ones. He says AI hasn't just changed the job market, it's inverted it. The people who used to be safe are now the ones at risk. "Everybody with the normal safe skills are dyslexics now, because the thing they can do that used to be valuable is not so valuable anymore." The skills that got you hired, reading, writing, regurgitating information are exactly what AI does cheapest and fastest. What's left is the thing that was never teachable in the first place. "If you actually have insights into anything and have real technical expertise, you can look at a company because you know something about how these things work." "I feel like Odin came down and said, you know what, you suffered so much as a kid, I'm just going to make the whole world so everyone else can suffer too. It's really an inversion." "The thing they need to learn to do is be more of an artist, look at things from a different direction, be able to build something unique." PS. We post daily content strictly for dedicated entrepreneurs, so if you’re one of them, make sure to follow us @entreprneursonx for more. Like and repost if you found value in this post:

Formal math may become the first domain where AI meaningfully exceeds elite human productivity because the environment perfectly suits machine iteration. Mathematics rewards patience, exhaustive search, symbolic manipulation, and immediate verification. Humans rely heavily on intuition because we are compute constrained. AI is not. The real breakthrough is not replacing mathematicians, but dramatically expanding how many problems humanity can realistically explore simultaneously together.

Google DeepMind's AlphaProof Nexus autonomously solved 9 open Erdős problems, some unsolved for 56 years, at a cost of a few hundred dollars per problem. It also proved 44 open OEIS conjectures, resolved a 15-year-old question in algebraic geometry, and discovered a novel algorithmic parameter in optimization theory that humans hadn't found. The core mechanism combines LLM reasoning (Gemini 3.1 Pro hype?!) with Lean formal verification. The AI generates proof attempts, Lean's compiler checks every logical step automatically. No human review needed to confirm correctness. The most surprising finding: a basic agent that simply alternates LLM generation with compiler feedback replicated all 9 Erdős successes. The full-featured system with evolutionary search and reinforcement learning only provided meaningful advantages on the hardest problems. This shows a more recent broader trend: as foundation models improve, simple agentic loops are catching up to complex specialized architectures . What sets this apart from OpenAI's informal proof approach: formal verification acts as an automatic filter. The failure analysis showed the AI frequently hallucinated lemmas it claimed were established results, and often disguised the core difficulty by rephrasing it as a helper lemma. Informal proofs would let these errors pass. Lean catches them immediately. The agent also detected misformalizations in existing mathematical literature, correcting ambiguities in problem statements before solving the corrected versions. It served as both a solver and a diagnostic tool. Current limitations are real. Successes cluster in combinatorics, number theory, and optimization where Lean's math library is mature. Problems requiring substantial new theory remain out of reach. Most Erdős problems still weren't solved tho.

What happened at Microsoft and Uber also exposes Goodhart’s Law in practice. Once AI usage became a performance signal, employees optimized for visible consumption rather than efficient deployment. Organizations accidentally incentivized token burn instead of measurable business value, creating runaway costs disconnected from actual productivity improvements or outcomes.

Microsoft just banned 100,000 engineers from using Claude Code. This wasn’t because it doesn’t work, but because the API bill was literally higher than the salaries it was meant to replace. Between this and uber burning their whole annual AI budget by april, the "AI will save us money" era might already be cooked.