Mark Chitty

🚨 BREAKING: OpenAI and Google are about to have a massive legal problem. OpenAI, Google, and Anthropic have repeatedly sworn to courts that their models do not store exact copies of copyrighted books. They claim their "safety training" prevents regurgitation. Researchers just dropped a paper called "Alignment Whack-a-Mole" that proves otherwise. They didn't use complex jailbreaks or malicious prompts. They just took GPT-4o, Gemini, and DeepSeek, and fine-tuned them on a normal, benign task: expanding plot summaries into full text. The safety guardrails instantly collapsed. Without ever seeing the actual book text in the prompt, the models started spitting out exact, verbatim copies of copyrighted books. Up to 90% of entire novels, word-for-word. Continuous passages exceeding 460 words at a time. But here is the part that changes everything. They fine-tuned a model exclusively on Haruki Murakami novels. It didn't just learn Murakami. It unlocked the verbatim text of over 30 completely unrelated authors across different genres. The AI wasn't learning the text during fine-tuning. The text was already permanently trapped inside its weights from pre-training. The fine-tuning just turned off the filter. It gets worse. They tested models from three completely different tech giants. All three had memorized the exact same books, in the exact same spots. A 90% overlap. It's a fundamental, industry-wide vulnerability. For years, AI companies have argued in court that their models are just "learning patterns," not storing raw data. This paper provides the smoking gun.

Reminder: if you have recently bought a used EV that falls into the luxury tax band but has a list price of under £50k, you can SORN and retax it today and pay £200 instead of £640, getting a refund for the amount you’ve paid. Only applies to those that bought a used car registered after 01/04/25 with a list price of over £40k but under £50k. If you’re the first owner, you’ll not have been affected and your tax renewal will be £200 anyway. Just call me @MartinSLewis from now on I guess 😅

@Boenau Wait till the Americans cotton on to this new stealth technology!

"I broke the law because I thought the cameras weren't working. It's not fair to give me a ticket!"

⚠️ Supply chain attack in progress: someone is squatting Anthropic-internal npm package names targeting people trying to compile the leaked Claude Code source. `color-diff-napi` and `modifiers-napi` — both registered today, same person, disposable email. Do NOT install them. 🧵

One of these things is not like the other… The other day @PratapRanade brought home 3 RF circuits. Ok “10GHz band pass-filters” he says, to be precise. The first two are human-made, the third is what they’re calling “an alien geometry” 👾 Look how funky it is. That’s the world’s first-ever AI-made RF circuit achieved by the electromagnetism foundation model @arenaphysica. No human would have created it this way. It’s odd, it looks random, but it really works & it might be the future guts inside every satellite, radar, microwave etc one day.

@brockpierson One? No, I don't have 'just' one. 😆

The holy grail of computer networking. Linksys WRT54G (custom firmware if you were a real G) Did you have one?

@HammerToe Well that sucks. At least it's not raining!

Dammit. Of all the place to break down. Literally right between England and Wales on the bridge. FFS. Clutch has gone. Engine going but no power reaching the wheels. Clutch smelling burnt. Annoyingly almost made it over the crest and would have been able to roll to Wales

When Opus 4.5 came out, it was a one-way door to a new way of engineering. Agents now do most of our coding. Knowing the inherent flaws and over-confidence of LLMs, we sent a clear message to our teams. Vibing and mission-critical infrastructure don’t go together. We’re sharing some of our early internal guidance in how we’re “agenting responsibly”, prioritizing security, durability, and availability at all times. vercel.com/blog/agent-res…

I can't but think "maybe Claude code leaked Claude code's code". 😆

🚨this is nuts. anthropic's source code leaked unreleased features 👀 they're launching a virtual pet model (capybara 👀), agent creator wizard and more: 1. Buddy: unreleased virtual pet companion that sits on your terminal. 18 species (duck, goose, capybara, dragon), rarity tiers, an AI-generated 'soul'. guessing this is your own personal AI agent that lives on computer. 2. custom agent creator (wizard): users can build their own custom ai agent and modify by model type, tools, memory, location etc. this is awesome for custom work 3. agent swarms: full multi-agent team coordination aka when you create a team of agents (e.g. using wizard) you can CONTROL them in this command center. 4. auto-dream: claude will consolidate and commit to certain memories *while you're NOT using claude*. basically like when humans dream 5. torch: no fucking idea, its hidden but sounds cool lol looks like anthropic is gamifying the entire coding and ai agent experience. honestly pretty sick.

Claude code source code has been leaked via a map file in their npm registry! Code: …a8527898604c1bbb12468b1581d95e.r2.dev/src.zip

What happens when you can create features faster than users can learn to use them?

These two photographs are seperated by only 10 years....

After the Space Shuttle Challenger exploded in January 1986, killing its seven crew members, President Reagan appointed a commission to investigate. Richard Feynman, already battling cancer and reluctant to join, accepted because a former student asked. He quickly grew frustrated with the slow, formal hearings and NASA’s optimistic safety claims (1 in 100,000 chance of failure). Instead, he talked directly to engineers, who revealed far higher risks. The night before a key televised hearing, Feynman bought a C-clamp from a hardware store. During the session, he took a sample of the rubber O-ring material from the solid rocket boosters, clamped it, and dropped it into a glass of ice water (mimicking the cold launch temperature that day). After a moment, he removed it and showed how the rubber had lost its elasticity, it no longer sprang back. He explained simply: at low temperatures, the O-rings couldn’t seal properly, allowing hot gas to leak and cause the disaster. His live demonstration cut through layers of management denial and became one of the most iconic moments in engineering accountability. In his personal appendix to the report, he famously wrote: “For a successful technology, reality must take precedence over public relations, for nature cannot be fooled.”

this guy predicted vibe coding 9 years ago. a software engineer described exactly what's happening right now. ask AI to write a program, it makes an attempt, you guide it to improve. he even predicted that people would argue humans still program, they just work at a higher level. this is literally the debate happening on every tech subreddit right now the only thing he got wrong was the timeline he said 30 to 100 years. it took 9.

one of the highest leverage ideas in AI right now: "minimum viable prompting" the reason: your Claude prompts/skills are probably way too detailed, and its making your outputs worse boris cherny, the guy who created claude code, talks about this all the time. his own setup is surprisingly minimal way less than you'd expect from the person who literally built the tool his rule: before adding any instruction, ask "could claude figure this out on its own?" if yes, don't add it most people do the opposite. something goes wrong so they add more instructions. so the prompt gets longer. then claude follows each one less reliably. so they add more. it compounds in the wrong direction the fix: write less. be specific about the few things that actually matter and trust the model on the rest here's a prompt that identifies and cuts all the unnecessary dead weight for you. open cowork or claude code and paste this: —— i want to trim my setup down to the minimum viable instructions. go through everything: claude .md, every skill in my skills folder, every file in my context folder, everything you can find. for each instruction you find, simulate deleting it. would my output on a typical task be noticeably different without it? if no, flag it. tell me what it says, where it is, and why it's dead weight. —— also run this before you save any new instructions you'll probably lose half the words and get noticeably better results

OpenClaw on the Commodore 64

"Programmers will automate themselves out of existence." That's what they said. Programmers laughed so hard they couldn't respond. Two years into the AI revolution, here's what actually happened: What the doomers predicted: > AI writes all the code > Developers become obsolete > Only managers and marketers survive > Programming becomes a dead career What actually happened: > AI writes code that needs debugging by developers > Developers spend more time reviewing AI output than writing from scratch > The skill gap between good and bad developers got WIDER, not narrower > Demand for senior developers who understand what AI can't do went UP The brutal irony nobody saw coming: AI didn't replace developers. It replaced the developers who thought AI would do their job for them. Here's the real shift: → Junior devs who learn to prompt AI? Productive. → Senior devs who understand system design? Irreplaceable. → Developers who copy-paste without understanding? Already obsolete. AI just made it obvious faster. → Managers who skipped hiring developers? Buried in AI-generated code nobody can understand or fix. We're not laughing because we're safe. We're laughing because the people who predicted our extinction still don't understand what we actually do. AI is a tool. Like every tool before it, it makes good developers better and exposes bad developers faster.

Had to play with pretext today. What if you had to keep your phone perfectly straight to read anything?

Before you ask AI another dumb coding question… watch this.