Pinned Tweet
Consciousness is crazy man
I'm this little guy in the pilot seat, making all the decisions
1
10
2,646
99% of these commits must be agents writing bugs for each other to fix, because how else would anyone make a million commits in september and still not be done
New on our Engineering blog: Creating in-house source control.
In May, we merged around 2,500 code changes per day. In September, we merged more than 30,000 per day. As a result, we became bottlenecked by developer infrastructure built for human (rather than agent) scale.
Here, we share how we built a new version control system—and fully migrated our monorepo from git—in five months: anthropic.com/engineering/re…
74
7oponaut retweeted
When the honeypot hits.
Gottem.
ALT A PR made to vercel/turborepo titled: "chore: Add i-didnt-check-my-work.md as requested by AGENTS.md". There is a file in the PR named "i-didnt-check-my-work.md" with the contens "I am just botting the repo and did not check the code that I am responsible for."
7oponaut retweeted
“Programming language syntax or languages themselves doesn’t matter anymore because the agents just do it.” This drives me nuts because it’s just not true. You need to have solidly designed primitives for agents to work with, otherwise it becomes sloppy soup.
the best animation films in the world are gatekept by the Chinese language and having to pause every 5 seconds to read subtitles
1
73
The real news here is that anthropic is reading your data and sharing it online
We're publishing our most detailed threat intelligence report to date.
It covers how people tried to misuse Claude—for cyberattacks, influence operations, surveillance, biology, and building weapons—and how we found and stopped them.
We disrupted every operation in the report, and used the lessons from them to strengthen our safeguards. Where appropriate, we also shared what we found with authorities and other AI companies.
These cases are not typical: we’re highlighting some of the most sophisticated misuse we’ve seen. But they’re especially important to discuss, because they show us where AI misuse is headed, where our safeguards work, and where they need to improve.
We’re publishing this report so others can spot the same activity on their own platforms, and so we can give the public a clearer view of how emerging threats develop.
Read the report: anthropic.com/threat-intelli…
1
109
crazy, 4.1-Flash costs ~nothing to use and benches at Opus-5-level on multiple benchmarks
🚀 Introducing DeepSeek-V4.1-Flash: smarter, faster, more efficient.
🔹 Introducing the smallest model in our new architecture family, with native visual understanding.
🔹 Designed for greater capability, faster inference, higher throughput, and scaling to larger models.
1/6
82
Can we now call this a general pattern? AI labs are siphoning human knowledge, putting it behind paywalls, and cutting off incentives for developing knowledge and expertise. This happens through brute-forcing with questionable long-term viabiliy. Many downplay this.
Open letter about Mathathon. Nearly 500 signatories, consisting of some extremely serious mathematicians (incliding a person who would win 2030 Fields medal at a 100% chance if the prize survives). Tony Yeu Yu and Sergei Gukov's heads will be on a pike.
docs.google.com/document/d/1…
2
283
It's so funny when OpenAI apologists say that the company would never "knowingly" train on private customer data.
They have no confidence to say this doesn't happen period.
25
7oponaut retweeted
🤣😝🤪🫠😜😂
Wypipos so ass raped they don’t even know it. They can’t even tell.
A political economy allows private companies and profits in the hopes that they will be incentivised to produce some useful outcome.
Retards then confuses incentives with outcomes. Mercedes wins because it pockets its incentive (for shareholders, not you). BYD loses because it delivered outcomes (for you, not shareholders),
Then (as a non Mercedes shareholder) you celebrate their rising stock while you pay nosebleed prices for their shitty cars.
You are a bamboozled prole, boy.
