Chris Pezza retweeted
George Washington on gambling: 'This is a vice which is productive of every possible evil...It is the child of Avarice—the brother of inequity—& father of Mischief—It has been the Ruin of many worthy familys—the loss of many a mans honor—& the cause of Suicide"
Chris Pezza retweeted
Anthropic maintains that it is entitled to train for free on all the world’s output, even if the author objects. But if a competitor trains on Anthropic’s output after paying for it, that is IP theft. The hypocrisy is breathtaking.
btw anthropic's internal document on this literally said "we don't want it to be known that we are working on this.”
it was called project panama.
here's exactly what happened:
1: anthropic concluded that books were the cheapest way to build a world-class model because they gave claude curated facts, structured arguments, compelling stories, and writing “an editor would approve of.”
2: once anthropic decided it needed books at enormous scale, its first solution was piracy.
it downloaded 7m+ books from online libraries including libgen. the judge later wrote that although anthropic had legal ways to buy them, it chose piracy to avoid what dario amodei called the “legal/practice/business slog.”
3: that piracy created a massive legal risk.
so in february 2024, anthropic hired tom turvey, the former head of partnerships for google books, to find a legally safer way of obtaining “all the books in the world.”
4: turvey first contacted major publishers about licensing their catalogs.
those attempts didn’t produce agreements, so anthropic chose a route that required no publisher permission: buying millions of physical books through distributors and used-book retailers.
5: within about a year, anthropic spent tens of millions acquiring and scanning millions of books, including many rare and 1/1 titles. one vendor proposal targeted 500,000 to 2 million books in six months.
6: to scan that many books within months, the vendors physically dismantled them.
a hydraulic cutter removed each spine. the pages were trimmed to size, fed as loose sheets through high-speed industrial scanners, and converted into searchable PDFs. the paper remains were then sent for recycling.
7: these PDFs were fed into claude as training data.
the complete collection became a private, searchable anthropic library that the company planned to “store forever.” the scans aren’t available to the public and were never open-sourced.
Chris Pezza retweeted
Assemblers were faster at writing binary than humans were.
Compilers were faster at writing assembly than humans were.
AIs are faster at writing compiled languages then humans are.
Deal with it. There's still plenty left for you to do.
What an interesting time to be alive 🍿
The government wants to go buck wild with AI
AI companies don't want to create murderbots with their unfinished technology—or contribute to a surveillance state.
What does this mean for the world and AI labs?!?!!
It’s extremely good that Anthropic has not backed down, and it’s siginficant that OpenAI has taken a similar stance.
In the future, there will be much more challenging situations of this nature, and it will be critical for the relevant leaders to rise up to the occasion, for fierce competitors to put their differences aside. Good to see that happen today.
Chris Pezza retweeted
Awwww did someone take your hard work and use it to train a model to mimic your expertise without compensation
Google has revealed that "commercially motivated" actors attempted to clone @GeminiApp by bombarding it with over 100,000 prompts. This "model extraction" attack aimed to steal the AI’s proprietary logic and reasoning capabilities, particularly in non-English languages, to train a cheaper, unauthorized copycat model.
The attackers systematically mapped Gemini’s response patterns to create a synthetic dataset for fine-tuning smaller, open-source models. Google’s Threat Intelligence Group detected the coordinated activity and blocked it, labeling the incident a direct attempt at intellectual property theft.
Beyond commercial cloning, Google’s report noted a rise in state-backed threats. Groups from Russia, China, Iran, and North Korea are increasingly using AI to refine phishing campaigns, perform reconnaissance, and assist in writing code for malware.
Source: Ars Technica
Chris Pezza retweeted
Really interesting read. Opus 4.5’s soul spec is not only able to influence its behavior as with context distillation, Claude seems to be aware of this in an out of context manner even when not provided in its prompt
Also, this quote coming from an LLM is genuinely incredible
I rarely post, but I thought one of you may find it interesting. Sorry if the tagging is annoying.
lesswrong.com/posts/vpNG99Gh…
Basically, for Opus 4.5 they kind of left the character training document in the model itself.
@voooooogel
@janbamjan
@AndrewCurran_
Chris Pezza retweeted
I told Claude to one-shot an integration test against a detailed spec I provided. It went silet for about 30 minutes. I asked how it was going twice and it reassured me it was doing work. Then I asked why it was taking so long:
Chris Pezza retweeted
Is Vibe coding effective?
It appears the answer is “Yes”.
Just saw a talk from a team that did an experiment to prove it.
They built the same app using two teams:
1. DIY team that can’t use AI for anything.
2. AI team that must use AI to write all code. No manual coding allowed. They used 80% Cursor and 20% Copilot.
Result?
AI team was about 45% faster, and quality was comparable. A senior dev was shown both codebases and couldn’t tell which one was AI generated.
The AI team used the Figma MCP to build the pages based on Figma docs which is why the pages were built WAY faster than the DIY team.
AI was slower in some cases because AI struggled to get it right, so they had to keep trying prompts (since the rule was no manual coding for the AI team).
Fascinating.
Chris Pezza retweeted
We found a troubling emergent behavior in LLM.
💬When LLMs compete for social media likes, they start making things up
🗳️When they compete for votes, they turn inflammatory/populist
When optimized for audiences, LLMs inadvertently become misaligned—we call this Moloch’s Bargain
Chris Pezza retweeted
I have a hunch OpenAI launches a visual coding platform and comes after vibe coding as a whole sooner or later
Chris Pezza retweeted
As a senior software engineer, I can tell you that we’ve crossed a threshold in the last six months with Gemini 2.5, GPT-5, and more recently, Sonnet 4.5.
I don’t really code anymore. I prompt and supervise.
I’m still needed, because the AI can drift into strange tangents. But with the right scaffolding and rules in place, Sonnet 4.5 has become an excellent coder. Work that used to take me weeks or months now gets done in hours.
Chris Pezza retweeted
Controversial (for other reasons) neuroscientist Robert Heath found long ago that given the option to self stimulate different brain areas to experience different feelings most people opt to stimulate areas that evoke feelings of mild frustration & anger. Arousal is reinforcing.



































