Writer at Silicon Canals

Machine Room

Contributor

Machine Room covers AI and deep tech: the builders, the breakthroughs, and the systems reshaping how technology gets made, with more interest in what ships than in what gets announced.

AI’s first visible employment shock may be hitting the bottom rung of the career ladder: by June 2026, employment among 22–25-year-olds in highly AI-exposed occupations was running 19% behind their less-exposed peers, largely because companies were hiring fewer juniors—not firing experienced workers—raising an uncomfortable question about who becomes tomorrow’s senior analyst, lawyer or coder.

Updated payroll data point to a widening early-career employment gap in AI-exposed work, driven more by reduced hiring than by layoffs.

Sep 1, 2026

Between May and December 2025, the share of American students from middle school through college using AI for homework rose from 48% to 62%. Yet 67% believed heavier AI use harms students’ critical thinking—and only about one-third said their school had established a schoolwide AI policy.

A RAND survey found rapid growth in students’ homework AI use alongside rising concern about critical thinking and a fragmented school policy landscape.

Sep 1, 2026

In January 2026, the largest human–AI creativity comparison yet tested language models against 100,000 people. GPT‑4 beat the average person at generating unexpected associations — but even the most creative half of humanity outperformed every model tested, with the top 10% far ahead.

GPT-4 beat the human average on a large divergent-association benchmark, but the upper half of the 100,000-person sample still outscored every model tested, revealing a very different human–AI creativity story in the tail.

Aug 26, 2026

One of the most counterintuitive findings from the first wave of AI productivity research is that the biggest gains often go to the people who were weakest to begin with — one study found productivity jumped 34% for novice workers while experienced employees barely improved, suggesting AI may compress skill advantages that once took years to build.

Early workplace studies suggest generative AI can sharply steepen the learning curve for novices, while offering much smaller gains to experts.

Aug 23, 2026

For decades, automation fears focused on factory workers while knowledge workers looked comparatively safe. AI’s trajectory has begun to flip that assumption: software developers, financial analysts and junior lawyers are already seeing parts of their work compressed or automated, while robots capable of broadly replacing human factory workers remain years behind.

Generative AI is compressing parts of software, finance and legal work before general-purpose factory robots have reached broad, economic deployment.

Aug 21, 2026

Google’s ATLAS study analyzed nearly 15 million AI interactions across more than 150 countries and found that, for the non-routine cognitive work where AI is used most heavily, fewer than 10% of conversations attempted to automate a task end-to-end. For now, the real-world picture looks much more like humans working with AI than AI simply replacing them.

Google’s ATLAS dataset suggests most observed cognitive work with AI is collaborative, but its scope and enterprise exclusions limit what the finding proves.

Aug 20, 2026

Technology

The Coldcard holders drained on 30 July had multiple safes, a safety deposit box and a device that never touched the internet, which is four defences against the same threat and none against the one that emptied them, a shape engineers call common cause failure

When every precaution points the same direction, they all fail together.

Aug 19, 2026

Technology

If you can’t keep your Bitcoin on an exchange that might freeze withdrawals, or in a hot wallet on a laptop that might get hacked, cold storage was the last answer anyone had left — and on 30 July roughly 1,082 coins drained out of devices that had never once been connected to the internet

A vulnerability in supposedly unhackable hardware wallets just moved $70 million in forty-one minutes, raising a terrifying question: if cold storage isn't safe, where is Bitcoin actually secure?

Aug 17, 2026

Technology

Epoch AI estimates that frontier models could fully use up the available stock of useful public human-written text sometime between 2026 and 2032. After that, developers may increasingly turn to private datasets and enormous quantities of AI-generated synthetic data — raising a strange possibility: a growing share of what future AI learns may ultimately have been written not by humans, but by earlier AI systems.

Epoch AI's 2026–2032 forecast is about effective public human text, not the end of new writing. The response may reshape data licensing, provenance and synthetic-data engineering.

Aug 14, 2026