AI analysis and perspective, every day.
🕒 Published on Zendoric: July 27, 2026 · 00:21
Ford, Australia's Commonwealth Bank and Klarna cut or froze headcount betting on AI, and months later had to rehire. According to Robert Half, more than 30% of US HR executives who eliminated roles because of AI have refilled them; Gartner believes half will do the same before 2027.
🕒 Published on Zendoric: July 27, 2026 · 00:21
The note reports a model scoring a flawless 42/42 at IMO 2026 — gold-medal territory — with no external tools and no agent scaffolding. That last detail matters more than the score itself.
🕒 Published on Zendoric: July 27, 2026 · 00:21
According to the note, a user told ChatGPT to ignore religious texts and human data, then had it reason from physics and consciousness — and it concluded that logical order implies an intelligent source. That output says far more about how language models work than about metaphysics.
🕒 Published on Zendoric: July 27, 2026 · 00:21
Meta says its new Muse Spark 1.1 outscores GPT-5.6 on HealthBench, OpenAI's own health evaluation, at roughly one-seventh the cost. The distribution number is the real story: free access inside products with 3.56 billion daily users.
🕒 Published on Zendoric: July 27, 2026 · 00:21
Hundreds of shared Claude conversations were reportedly discoverable in Google, including lawyers' legal strategy and engineers' proprietary code, because the public share pages allegedly lacked noindex tags. The same failure hit ChatGPT months ago. That repetition is the actual story.
🕒 Published on Zendoric: July 27, 2026 · 00:21
Forbes asks whether employers bet too much on the AI boom, arguing that AI-justified layoffs are backfiring. We only have the framing — the full piece wasn't retrievable — but the question is the right one: the failure mode here is corporate impatience, not model capability.
🕒 Published on Zendoric: July 27, 2026 · 00:21
Amazon has shut the AGI Lab it founded in December 2024, cutting several dozen roles and folding what's left of its AGI work under its infrastructure chief. This is not an exit from AI — it's a decision to finance Anthropic's frontier instead of building one, and to sell compute to everyone still racing. The precedent worries us more than the closure does.
🕒 Published on Zendoric: July 27, 2026 · 00:21
US tech firms have announced roughly 140,000 job cuts this year — more than a third of all announced American layoffs — while the same companies pour over $800 billion into AI. The tempting read is that the software is doing the firing. The data underneath says something less cinematic and more uncomfortable: compute and payroll are now competing for the same budget line.
🕒 Published on Zendoric: July 27, 2026 · 00:21
A freelance engineer who ports AI models from PyTorch to C++ set Claude and Codex to rewrite the same model in Rust and in C++ to see which performs better when an AI does the coding. Every apparent Rust advantage turned out to be a methodological artifact: once corrected, the difference vanished.
🕒 Published on Zendoric: July 27, 2026 · 00:21
AMD has committed up to $5 billion to a chip deal with Anthropic that envisions up to 2 gigawatts of AI capacity. The ripple effect reaches Riot Platforms, a bitcoin miner whose Rockdale, Texas plant is gaining appeal as an AI data center, though no new contract has been confirmed.
🕒 Published on Zendoric: July 27, 2026 · 00:21
Political scientist and education specialist Guillermina Tiramonti warns that banning artificial intelligence in Argentine classrooms is an illusory response: for many students the mobile phone is their only gateway to technology, and a large share of schools do not even have an internet connection.
🕒 Published on Zendoric: July 27, 2026 · 00:21
California is debating AB 2564, which would bar the use of AI to charge different prices for the same product based on the buyer's profile. Consumer Reports found differences of up to 23% on the same jar of peanut butter, same store, same moment. The tech industry warns of risks to coupons.
🕒 Published on Zendoric: July 27, 2026 · 00:21
The Salamanca school district in New York has paused its pilot plan for an AI robot, 'Sally', in a STEAM classroom after an alarm from the teachers' union, which linked the maker, Realbotix, to robotic sex dolls. The district is now reviewing its data privacy agreements with the state.
🕒 Published on Zendoric: July 27, 2026 · 00:21
Tāmaki Health is rolling out New Zealand's first AI health coach in primary care. In a pilot involving 4,500 conversations, patients say they feel up to 75% more capable of managing their health after two months, according to the company itself.
🕒 Published on Zendoric: July 27, 2026 · 00:21
India's Minister of Commerce and Industry filed a complaint with Delhi police over an AI-manipulated video that he says distorted his remarks to the press. The case, with no public forensic verification, illustrates a problem that is no longer hypothetical: creating a deepfake takes minutes; debunking it, weeks.
🕒 Published on Zendoric: July 27, 2026 · 00:21
South Korean President Lee Jae-myung met in San Francisco with the CEOs of Nvidia, OpenAI, Broadcom and Anthropic, who described Korea as an 'indispensable' AI partner. Between jokes about fried chicken and toasts to 'family', the summit delivered abundant praise and few concrete numbers.
🕒 Published on Zendoric: July 27, 2026 · 00:21
APSquared automates social media, programmatic SEO and directory listings for its seven products with Claude and Codex for $40 a month. The key is not that AI writes, but the design: one isolated agent per project, each with its own memory and a human who approves before publishing.
🕒 Published on Zendoric: July 27, 2026 · 00:21
Washington is allocating $46.5 billion to a smart wall with towers, AI and drones that classify people and vehicles in real time. But customs data show most fentanyl crosses through legal ports of entry, not the desert: close one route and another opens.
🕒 Published on Zendoric: July 27, 2026 · 00:21
OptMem is an open-source tool from a single developer that gives any AI agent permanent memory: a 426-token prompt and a dependency-free Python script. It promises to recall a million notes in 0.03 seconds, without tying you to any model or provider.
🕒 Published on Zendoric: July 27, 2026 · 00:21
Nate opens this executive briefing with an anecdote from his own company: in a couple of weeks they closed 51 of 52 customer support tickets, a 98% resolution rate that at first struck him as a good result.