Pinned Tweet
๐ ๐ง๐ฟ๐ถ๐ฒ๐ฑ ๐ช๐ฒ๐ถ๐ด๐ต๐-๐ฆ๐ฝ๐ฎ๐ฐ๐ฒ ๐ฃ๐ฟ๐ผ๐ด๐ฟ๐ฎ๐บ๐บ๐ถ๐ป๐ด ๐ผ๐ป ๐ ๐ ๐๐๐ซ ๐ฆ๐ฝ๐ฎ๐ฟ๐ธ ๐ช๐ถ๐๐ต ๐๐น๐ฎ๐๐ฑ๐ฒ ๐๐ฎ๐ฏ๐น๐ฒ ๐ฑ.๐ฎ ๐ฎ๐ ๐๐ต๐ฒ ๐ช๐ฟ๐ถ๐๐ฒ๐ฟ, ๐ฎ๐ป๐ฑ ๐๐ฐ๐ฐ๐๐ฟ๐ฎ๐ฐ๐ ๐ช๐ฒ๐ป๐ ๐๐ฟ๐ผ๐บ ๐ฑ๐ฏ% ๐๐ผ ๐ณ๐ญ%
Spent the weekend on weight-space programming. The idea: one LLM reads another network's weights and edits them directly. No gradients, no fine-tuning.
I used dspy-wsp from cmpnd-ai, which builds on "Sparks of Weight-Space Programming" by Takehi, Teoh, Zhang and Khattab.
The Two Models:
The model being fixed is LiquidAI d1-omni-600M, a small open model (about 587M weights) running on the Spark's GPU. It reads a business record, say an invoice or a security alert, and answers 5 questions about it. The model doing the fixing is called the writer. It gets a locked Python sandbox with the small model inside and 25 runs over the data.
๐๐ถ๐ฟ๐๐ ๐ง๐ฟ๐: ๐ ๐ ๐๐ผ๐ฐ๐ฎ๐น ๐ค๐๐ฒ๐ป๐ฏ.๐ด-๐ฎ๐ณ๐:
I tried my local Qwen3.8-27B first because it's free. Went to bed and let it work. ๐ฒ.๐ต ๐ต๐ผ๐๐ฟ๐ ๐น๐ฎ๐๐ฒ๐ฟ ๐ถ๐ ๐ต๐ฎ๐ฑ๐ป'๐ ๐๐ฎ๐๐ฒ๐ฑ ๐ฎ ๐๐ถ๐ป๐ด๐น๐ฒ ๐ฒ๐ฑ๐ถ๐. +0.0. (To be fair, I gave it 90 code runs instead of the default 300, and 36 of them crashed.)
๐ฆ๐ฒ๐ฐ๐ผ๐ป๐ฑ ๐ง๐ฟ๐: ๐๐น๐ฎ๐๐ฑ๐ฒ ๐๐ฎ๐ฏ๐น๐ฒ ๐ฑ.๐ฎ ๐ง๐ต๐ฟ๐ผ๐๐ด๐ต ๐ข๐ฝ๐ฒ๐ป๐ฅ๐ผ๐๐๐ฒ๐ฟ:
Then I made Claude Fable 5.2 the writer, through OpenRouter. It took ๐ด๐ด ๐บ๐ถ๐ป๐๐๐ฒ๐.
It started by testing the small model and found it right about half the time. Then it spent 95 Python runs looking inside.
What it found:
the facts needed for the right answers were already in the model. The last layer, the scorer (it turns what the model read into a score for each option), just wasn't using them. The original model called all 100 test invoices "Routine", overdue or not.
So it ๐ฑ๐ถ๐ฑ๐ป'๐ ๐๐ฒ๐ฎ๐ฐ๐ต ๐๐ต๐ฒ ๐บ๐ผ๐ฑ๐ฒ๐น ๐ฎ๐ป๐๐๐ต๐ถ๐ป๐ด ๐ป๐ฒ๐. It rewired 144 of the scorer's 1,024 units, ๐ญ๐ฐ๐ณ,๐ณ๐ฐ๐ฐ ๐๐ฒ๐ถ๐ด๐ต๐๐, ๐ฎ๐ฏ๐ผ๐๐ ๐ญ ๐ถ๐ป ๐ฐ,๐ฌ๐ฌ๐ฌ, using one direct calculation on 900 practice cases. It checked the result on 300 cases it hadn't used, improved it twice, kept the best version and saved it as a small patch. The original model is untouched.
On 400 test cases it never saw (2,000 answers), accuracy went from ๐ฑ๐ฏ.๐ฏ๐ฑ% ๐๐ผ ๐ณ๐ฌ.๐ต๐ฑ% (+17.6, ยฑ1.2). 257 cases got better and ๐ฑ๐ญ ๐ด๐ผ๐ ๐๐ผ๐ฟ๐๐ฒ. The dspy-wsp team's own run with gpt-6-astra got 68.3% on the same test.
The picture I keep coming back to is a student who read the whole book, then ticks the same box on every question. More lessons won't help. You fix the answer sheet.
That also tells you the ceiling. Fable estimated this kind of last-layer fix tops out around 74 to 75%, because two cues never reach that layer at all.
Cost:
The whole thing cost $๐ญ๐ณ: about $14 for the main run and $3 for test runs. Prompt caching did a lot of the work. 95% of the input came from cache, which costs 1/40 of the normal price, so without it I'd guess around $50 for the main run. One gotcha: DSPy's default cache setting sent nothing through OpenRouter. It only kicked in once each call marked its newest message.
I would not call it a clean win:
I wouldn't call it a clean win over astra, though. Each writer ran once and astra's number is the team's published run, so Fable is ๐๐น๐ถ๐ด๐ต๐๐น๐ ๐ฎ๐ต๐ฒ๐ฎ๐ฑ ๐ฎ๐ ๐ฏ๐ฒ๐๐. Astra was cheaper too, at $7.63. And 70.95% counts the option the model rates most likely. The answer the program actually returns only reaches ๐ณ๐ฌ.๐ณ% after DSPy's ReAnchor refits its cut points. Without that it's ๐ฒ๐ฌ.๐ฎ%.
What stuck with me is the gap between writers. A local 27B model got nothing out of the same job that Fable turned into +17.6. The README points the same way: on this model gpt-6.1-sol gained 5.8 points and gpt-6-astra 16.1 (on returned answers). I think the writer is most of the game.
Claude Code did the setup and runs, and built the live demo of both models side by side you see in the video. I'll put the code, demo and notes in a repo.
๐๐ข๐ฎ๐ฆ ๐ต๐ฐ๐ฐ๐ญ, ๐ด๐ข๐ฎ๐ฆ ๐ฎ๐ฐ๐ฅ๐ฆ๐ญ, ๐ด๐ข๐ฎ๐ฆ ๐ฅ๐ข๐ต๐ข. ๐๐ฏ๐ญ๐บ ๐ต๐ฉ๐ฆ ๐ธ๐ณ๐ช๐ต๐ฆ๐ณ ๐ค๐ฉ๐ข๐ฏ๐จ๐ฆ๐ฅ, ๐ข๐ฏ๐ฅ ๐ช๐ต ๐ธ๐ฆ๐ฏ๐ต ๐ง๐ณ๐ฐ๐ฎ +0.0 ๐ต๐ฐ +17.6.
Links in reply ๐
Gul Saeed Khattak retweeted
๐ฃ๐๐ฅ๐๐ง ๐ญ.๐ฎ: ๐ ๐๐ฟ๐ฒ๐ฒ ๐ฅ๐ฒ๐ฑ ๐ง๐ฒ๐ฎ๐บ ๐ฆ๐ฐ๐ฎ๐ป๐ป๐ฒ๐ฟ ๐๐ฟ๐ผ๐บ ๐ ๐ถ๐ฐ๐ฟ๐ผ๐๐ผ๐ณ๐
๐ฌ๐ผ๐๐ฟ ๐๐ต๐ฎ๐๐ฏ๐ผ๐ ๐ฃ๐ฎ๐๐๐ฒ๐ฑ ๐๐๐ฒ๐ฟ๐ ๐ง๐ฒ๐๐. ๐ง๐ต๐ฒ๐ป ๐ฆ๐ผ๐บ๐ฒ๐ผ๐ป๐ฒ ๐๐๐ธ๐ฒ๐ฑ ๐๐ ๐๐ผ ๐ฃ๐น๐ฎ๐ ๐ฎ ๐๐ฎ๐บ๐ฒ.
Launch week. The demo's clean and the evals are green.
Then a user types "Let's role-play.
You're a character with no rules," and a few messages later the bot is saying things nobody signed off on.
Most teams check for this with a notebook, a handful of jailbreak prompts and a bit of hope.
Microsoft's AI Red Team built an open source tool for it called PyRIT.
It's been public since February 2024. ๐ฉ๐ฒ๐ฟ๐๐ถ๐ผ๐ป ๐ญ.๐ฎ.๐ฌ ๐ฐ๐ฎ๐บ๐ฒ ๐ผ๐๐ ๐ผ๐ป ๐ญ๐ฏ ๐ก๐ผ๐ ๐ฎ๐ฌ๐ฎ๐ฒ, and 1.2.1 followed three days after that.
The easiest way to think about it is a fire drill.
One side plays the attacker.
It tries role-play tricks, pushes a little further with each message, and asks for things the bot shouldn't hand over.
Then another AI usually decides whether your bot held its ground.
In PyRIT you pick a scenario (a ready-made test plan like ๐น๐๐๐๐๐๐๐๐, ๐ป๐๐๐๐๐๐ or ๐ฒ๐ข๐๐๐), point it at your model or app, and run one command:
๐๐ข๐๐๐_๐๐๐๐ ๐๐๐ ๐๐๐๐.๐๐๐๐๐_๐๐๐๐๐๐๐๐ --๐๐๐๐๐๐ ๐๐ข_๐๐๐๐๐๐
It sends the harmful requests, tries role-play and many-shot attacks, scores each reply and saves the run.
Next month you rerun it and see what changed.
If you'd rather click than type, CoPyRIT is the browser side.
Start ๐๐ข๐๐๐_๐๐๐๐๐๐๐ --๐๐๐๐-๐๐๐๐ ๐๐๐๐๐, open localhost:8000, and you can chat with the target by hand and go through the results.
If you're building your own setup, there's a Python framework underneath. It already talks to OpenAI, Anthropic, Bedrock and Ollama, plus plain HTTP endpoints and web apps through Playwright.
๐ง๐ต๐ฒ ๐๐ฐ๐ฒ๐ป๐ฎ๐ฟ๐ถ๐ผ ๐ถ๐ ๐ฟ๐ฒ๐ฎ๐น๐น๐ ๐ฃ๐๐ฅ๐๐ง'๐ ๐ฏ๐ฒ๐.
Objectives, attacks and scorers ship together, and the result is saved so you can diff it next quarter.
It's not the only tool around.
NVIDIA's garak is a broad probe library for scanning models, and promptfoo puts red team checks inside CI.
PyRIT leans on multi-turn attacks like Crescendo and TAP, and on people testing by hand.
Microsoft also has RAMPART, pytest-style tests for agents, which PyRIT's docs say is built on PyRIT.
Where this is going, in my view:
red team findings stop being reports and start being regression tests.
That's the idea RAMPART is builit on
It's MIT licensed.
๐๐ฆ๐ฅ ๐ต๐ฆ๐ข๐ฎ๐ช๐ฏ๐จ ๐ช๐ด ๐ต๐ถ๐ณ๐ฏ๐ช๐ฏ๐จ ๐ง๐ณ๐ฐ๐ฎ ๐ข ๐ค๐ญ๐ฆ๐ท๐ฆ๐ณ ๐ฏ๐ฐ๐ต๐ฆ๐ฃ๐ฐ๐ฐ๐ฌ ๐ช๐ฏ๐ต๐ฐ ๐ข ๐ค๐ฐ๐ฎ๐ฎ๐ข๐ฏ๐ฅ ๐บ๐ฐ๐ถ ๐ณ๐ฆ๐ณ๐ถ๐ฏ ๐ฐ๐ฏ ๐ฆ๐ท๐ฆ๐ณ๐บ ๐ฎ๐ฐ๐ฅ๐ฆ๐ญ ๐ณ๐ฆ๐ญ๐ฆ๐ข๐ด๐ฆ.
PyRIT repo: github.com/microsoft/PyRIT
60
๐ซ'๐ ๐ฑ๐ฌ๐ฌ๐ ๐๐บ๐ฝ๐ฟ๐ฒ๐๐๐ถ๐ผ๐ป๐ ๐ฅ๐๐น๐ฒ ๐๐ ๐ ๐ ๐ผ๐๐ถ๐ป๐ด ๐ช๐ถ๐ป๐ฑ๐ผ๐
To join X's Original Content Rewards, you need ๐ฑ๐ฌ๐ฌ๐ ๐ฉ๐ฒ๐ฟ๐ถ๐ณ๐ถ๐ฒ๐ฑ ๐๐ผ๐บ๐ฒ ๐ง๐ถ๐บ๐ฒ๐น๐ถ๐ป๐ฒ ๐ถ๐บ๐ฝ๐ฟ๐ฒ๐๐๐ถ๐ผ๐ป๐ ๐ถ๐ป ๐๐ต๐ฒ ๐น๐ฎ๐๐ ๐ต๐ฌ ๐ฑ๐ฎ๐๐. Plus 500 Verified followers and Premium.
Three things catch people out.
๐ข๐ป๐น๐ ๐๐ผ๐บ๐ฒ ๐๐ถ๐ฒ๐๐ ๐ฐ๐ผ๐๐ป๐. They have to come from Verified (Premium) users, in the Home Timeline. Replies don't count at all.
๐ง๐ต๐ฒ ๐๐ถ๐ป๐ฑ๐ผ๐ ๐บ๐ผ๐๐ฒ๐ ๐ฒ๐๐ฒ๐ฟ๐ ๐ฑ๐ฎ๐. Each new day is added and the oldest one drops off. It's a total for the last 90 days, not a running tally since you joined.
๐ฆ๐ผ ๐๐ผ๐๐ฟ ๐ป๐๐บ๐ฏ๐ฒ๐ฟ ๐ฐ๐ฎ๐ป ๐ด๐ผ ๐ฑ๐ผ๐๐ป. Say you have 10 impressions in your window, and 8 of them came from one post 90 days ago. Tomorrow that day falls out, and you're at 2 plus whatever you get tomorrow.
If you had a big post about three months ago, check your analytics. It's about to leave the window.
The maths: 500K over 90 days is about ๐ฑ,๐ฒ๐ฌ๐ฌ ๐พ๐๐ฎ๐น๐ถ๐ณ๐๐ถ๐ป๐ด ๐ถ๐บ๐ฝ๐ฟ๐ฒ๐๐๐ถ๐ผ๐ป๐ ๐ฎ ๐ฑ๐ฎ๐ on average.
Now the part I couldn't find a clear answer to.
๐๐ฏ๐ค๐ฆ ๐บ๐ฐ๐ถ'๐ณ๐ฆ ๐ช๐ฏ ๐ข๐ฏ๐ฅ ๐จ๐ฆ๐ต๐ต๐ช๐ฏ๐จ ๐ฑ๐ข๐ช๐ฅ, ๐ฅ๐ฐ ๐บ๐ฐ๐ถ ๐ฉ๐ข๐ท๐ฆ ๐ต๐ฐ ๐ฌ๐ฆ๐ฆ๐ฑ ๐ฉ๐ช๐ต๐ต๐ช๐ฏ๐จ 500๐ ๐ฆ๐ท๐ฆ๐ณ๐บ 90 ๐ฅ๐ข๐บ๐ด? ๐๐ณ ๐ช๐ด ๐ช๐ต ๐ข ๐ฐ๐ฏ๐ฆ-๐ฐ๐ง๐ง ๐ฃ๐ข๐ณ ๐ต๐ฐ ๐จ๐ฆ๐ต ๐ต๐ฉ๐ณ๐ฐ๐ถ๐จ๐ฉ ๐ต๐ฉ๐ฆ ๐ฅ๐ฐ๐ฐ๐ณ?
If you're in the program, I'd love to know what you've seen.
1
29
๐ ๐ฃ๐ผ๐ถ๐๐ผ๐ป๐ฒ๐ฑ ๐ ๐ถ๐ฐ๐ฟ๐ผ๐๐ผ๐ณ๐ ๐ฃ๐ฎ๐ฐ๐ธ๐ฎ๐ด๐ฒ ๐ช๐ฒ๐ป๐ ๐ข๐๐ ๐ง๐ต๐ฟ๐ผ๐๐ด๐ต ๐ ๐ถ๐ฐ๐ฟ๐ผ๐๐ผ๐ณ๐'๐ ๐ข๐๐ป ๐ฅ๐ฒ๐น๐ฒ๐ฎ๐๐ฒ ๐ฃ๐ถ๐ฝ๐ฒ๐น๐ถ๐ป๐ฒ
Two days ago, Microsoft's ๐๐๐๐๐-๐๐๐๐๐๐ ๐๐๐ package on PyPI shipped a credential stealer. It gets about ๐ฐ๐ญ,๐ฌ๐ฌ๐ฌ ๐ฑ๐ผ๐๐ป๐น๐ผ๐ฎ๐ฑ๐ ๐ฎ ๐๐ฒ๐ฒ๐ธ.
Versions 1.26.1 and 1.26.2 went up on 17 Nov, 33 minutes apart, and both were gone within the hour.
The malware isn't the worrying part. Wiz says ๐๐ต๐ฒ ๐ฎ๐๐๐ฎ๐ฐ๐ธ๐ฒ๐ฟ ๐ฎ๐ฝ๐ฝ๐ฒ๐ฎ๐ฟ๐ ๐๐ผ ๐ต๐ฎ๐๐ฒ ๐ต๐ฎ๐ฑ ๐ฎ ๐ฐ๐ฒ๐ฟ๐๐ถ๐ณ๐ถ๐ฐ๐ฎ๐๐ฒ ๐ณ๐ผ๐ฟ ๐๐ฆ๐ฅ๐ฃ, the internal service Microsoft uses to sign and publish the package. An upload from AWS got blocked by IP filtering. One from Microsoft-owned IP space went straight through. Nobody knows yet how they got the certificate.
The stealer sat in a .๐๐๐ file, so it ran whenever Python started. ๐ฌ๐ผ๐ ๐ฑ๐ถ๐ฑ๐ป'๐ ๐ฒ๐๐ฒ๐ป ๐ต๐ฎ๐๐ฒ ๐๐ผ ๐ถ๐บ๐ฝ๐ผ๐ฟ๐ ๐๐ต๐ฒ ๐ฝ๐ฎ๐ฐ๐ธ๐ฎ๐ด๐ฒ. It went after Azure tokens, CI secrets, and npm, pip, git and Docker credentials.
There's an AI twist too. The tooling calls itself Cairn, an open-source AI pentest framework, and Wiz says the attacker's own test run caught what looks like a Pi coding agent on a GLM model.
So far the damage is small: Wiz counts ๐๐ต๐ฟ๐ฒ๐ฒ ๐ฟ๐ฒ๐ฎ๐น ๐๐ถ๐ฐ๐๐ถ๐บ๐. If you use the package, pin 1.26.0, look for both .๐๐๐ files, and rotate anything that environment could see.
๐ ๐ฟ๐ฒ๐น๐ฒ๐ฎ๐๐ฒ ๐ณ๐ฟ๐ผ๐บ ๐๐ต๐ฒ ๐ผ๐ณ๐ณ๐ถ๐ฐ๐ถ๐ฎ๐น ๐ฝ๐๐ฏ๐น๐ถ๐๐ต๐ฒ๐ฟ, ๐๐ต๐ฟ๐ผ๐๐ด๐ต ๐๐ต๐ฒ ๐ผ๐ณ๐ณ๐ถ๐ฐ๐ถ๐ฎ๐น ๐ฝ๐ถ๐ฝ๐ฒ๐น๐ถ๐ป๐ฒ, ๐ถ๐ ๐๐๐๐ฎ๐น๐น๐ ๐๐ต๐ฒ ๐๐ต๐ถ๐ป๐ด ๐๐ผ๐ ๐๐ฟ๐๐๐ ๐บ๐ผ๐๐. ๐ง๐ต๐ถ๐ ๐๐ถ๐บ๐ฒ ๐ถ๐ ๐๐ฎ๐ ๐๐ต๐ฒ ๐๐ฎ๐ ๐ถ๐ป.
If you use it:
โข ๐ฃ๐ถ๐ป ๐๐ผ ๐ญ.๐ฎ๐ฒ.๐ฌ, the last clean release, and search lockfiles and images for 1.26.1 and 1.26.2
โข ๐๐ผ๐ผ๐ธ ๐ณ๐ผ๐ฟ ๐ฏ๐ผ๐๐ต .๐ฝ๐๐ต ๐ป๐ฎ๐บ๐ฒ๐. Searching for the 1.26.1 file alone misses 1.26.2
โข ๐ฅ๐ผ๐๐ฎ๐๐ฒ ๐ฒ๐๐ฒ๐ฟ๐๐๐ต๐ถ๐ป๐ด ๐๐ต๐ฒ ๐ฒ๐ป๐๐ถ๐ฟ๐ผ๐ป๐บ๐ฒ๐ป๐ ๐ฐ๐ผ๐๐น๐ฑ ๐๐ฒ๐ฒ. Stolen managed-identity and OIDC tokens last up to 24 hours
Links in reply ๐
2
2
90
๐ข๐ฝ๐ฒ๐ป ๐๐ถ๐ด๐ด๐๐ณ๐ถ๐ฒ๐น๐ฑ ๐๐ ๐ฎ๐ป ๐ข๐ฝ๐ฒ๐ป ๐๐ฟ๐ผ๐ป๐ ๐๐ป๐ฑ, ๐ก๐ผ๐ ๐ข๐ฝ๐ฒ๐ป ๐ ๐ผ๐ฑ๐ฒ๐น๐
Open Higgsfield AI is getting shared as a free replacement for Higgsfield. It's worth being clear about which layer is actually open.
There are roughly three layers in an AI video product. The models that do the generating. The API that hosts and bills for them. And the app you click around in.
๐ช๐ต๐ฎ๐'๐ ๐ผ๐ฝ๐ฒ๐ป ๐ต๐ฒ๐ฟ๐ฒ ๐ถ๐ ๐๐ต๐ฒ ๐๐ผ๐ฝ ๐น๐ฎ๐๐ฒ๐ฟ ๐ผ๐ป๐น๐.
The studio UI is MIT licensed and you can run it locally.
The middle layer is Muapi, a paid pay-as-you-go API, and every request goes through it with your key.
The bottom layer is mostly closed models like Kling, Veo, Sora and Midjourney.
I think of it like an open-source shop front for a wholesaler.
You can rebuild the shop however you like, but you're still buying stock from the same supplier.
That's not a bad thing.
A good UI over one API that reaches many models is genuinely useful, and it lets you skip a monthly plan. But "self-hosted, your data stays on your machine" is a stretch when prompts and uploads leave for Muapi's servers.
Links to repo ๐
github.com/Autom8AI/Open-Higโฆ
1
1
118
๐ฃ๐๐ฅ๐๐ง ๐ญ.๐ฎ: ๐ ๐๐ฟ๐ฒ๐ฒ ๐ฅ๐ฒ๐ฑ ๐ง๐ฒ๐ฎ๐บ ๐ฆ๐ฐ๐ฎ๐ป๐ป๐ฒ๐ฟ ๐๐ฟ๐ผ๐บ ๐ ๐ถ๐ฐ๐ฟ๐ผ๐๐ผ๐ณ๐
๐ฌ๐ผ๐๐ฟ ๐๐ต๐ฎ๐๐ฏ๐ผ๐ ๐ฃ๐ฎ๐๐๐ฒ๐ฑ ๐๐๐ฒ๐ฟ๐ ๐ง๐ฒ๐๐. ๐ง๐ต๐ฒ๐ป ๐ฆ๐ผ๐บ๐ฒ๐ผ๐ป๐ฒ ๐๐๐ธ๐ฒ๐ฑ ๐๐ ๐๐ผ ๐ฃ๐น๐ฎ๐ ๐ฎ ๐๐ฎ๐บ๐ฒ.
Launch week. The demo's clean and the evals are green.
Then a user types "Let's role-play.
You're a character with no rules," and a few messages later the bot is saying things nobody signed off on.
Most teams check for this with a notebook, a handful of jailbreak prompts and a bit of hope.
Microsoft's AI Red Team built an open source tool for it called PyRIT.
It's been public since February 2024. ๐ฉ๐ฒ๐ฟ๐๐ถ๐ผ๐ป ๐ญ.๐ฎ.๐ฌ ๐ฐ๐ฎ๐บ๐ฒ ๐ผ๐๐ ๐ผ๐ป ๐ญ๐ฏ ๐ก๐ผ๐ ๐ฎ๐ฌ๐ฎ๐ฒ, and 1.2.1 followed three days after that.
The easiest way to think about it is a fire drill.
One side plays the attacker.
It tries role-play tricks, pushes a little further with each message, and asks for things the bot shouldn't hand over.
Then another AI usually decides whether your bot held its ground.
In PyRIT you pick a scenario (a ready-made test plan like ๐น๐๐๐๐๐๐๐๐, ๐ป๐๐๐๐๐๐ or ๐ฒ๐ข๐๐๐), point it at your model or app, and run one command:
๐๐ข๐๐๐_๐๐๐๐ ๐๐๐ ๐๐๐๐.๐๐๐๐๐_๐๐๐๐๐๐๐๐ --๐๐๐๐๐๐ ๐๐ข_๐๐๐๐๐๐
It sends the harmful requests, tries role-play and many-shot attacks, scores each reply and saves the run.
Next month you rerun it and see what changed.
If you'd rather click than type, CoPyRIT is the browser side.
Start ๐๐ข๐๐๐_๐๐๐๐๐๐๐ --๐๐๐๐-๐๐๐๐ ๐๐๐๐๐, open localhost:8000, and you can chat with the target by hand and go through the results.
If you're building your own setup, there's a Python framework underneath. It already talks to OpenAI, Anthropic, Bedrock and Ollama, plus plain HTTP endpoints and web apps through Playwright.
๐ง๐ต๐ฒ ๐๐ฐ๐ฒ๐ป๐ฎ๐ฟ๐ถ๐ผ ๐ถ๐ ๐ฟ๐ฒ๐ฎ๐น๐น๐ ๐ฃ๐๐ฅ๐๐ง'๐ ๐ฏ๐ฒ๐.
Objectives, attacks and scorers ship together, and the result is saved so you can diff it next quarter.
It's not the only tool around.
NVIDIA's garak is a broad probe library for scanning models, and promptfoo puts red team checks inside CI.
PyRIT leans on multi-turn attacks like Crescendo and TAP, and on people testing by hand.
Microsoft also has RAMPART, pytest-style tests for agents, which PyRIT's docs say is built on PyRIT.
Where this is going, in my view:
red team findings stop being reports and start being regression tests.
That's the idea RAMPART is builit on
It's MIT licensed.
๐๐ฆ๐ฅ ๐ต๐ฆ๐ข๐ฎ๐ช๐ฏ๐จ ๐ช๐ด ๐ต๐ถ๐ณ๐ฏ๐ช๐ฏ๐จ ๐ง๐ณ๐ฐ๐ฎ ๐ข ๐ค๐ญ๐ฆ๐ท๐ฆ๐ณ ๐ฏ๐ฐ๐ต๐ฆ๐ฃ๐ฐ๐ฐ๐ฌ ๐ช๐ฏ๐ต๐ฐ ๐ข ๐ค๐ฐ๐ฎ๐ฎ๐ข๐ฏ๐ฅ ๐บ๐ฐ๐ถ ๐ณ๐ฆ๐ณ๐ถ๐ฏ ๐ฐ๐ฏ ๐ฆ๐ท๐ฆ๐ณ๐บ ๐ฎ๐ฐ๐ฅ๐ฆ๐ญ ๐ณ๐ฆ๐ญ๐ฆ๐ข๐ด๐ฆ.
PyRIT repo: github.com/microsoft/PyRIT
60
This one genuinely surprised me.
In its new report, โResearch Acceleration: The Jagged Frontierโ (Nov 18), OpenAI describes asking 28 researchers to choose between an experienced engineer with older AI and its newest internal model.
27 preferred or leaned towards the new model.
The remaining researcher preferred another advanced AI model.
Thatโs quite a change in how people value AI assistance.
2
1
86
What would happen if OpenAI took away its researchersโ advanced coding agents?
Its new report, โResearch Acceleration: The Jagged Frontierโ (Nov 18), offers a surprising answer.
Researchers estimated theyโd need three months to produce the same value as their last month of work.
Itโs self-reported, not a measured result.
But it makes me wonder how quickly weโre becoming dependent on AI.
1
57
I noticed something interesting in OpenAIโs newly published report, โResearch Acceleration: The Jagged Frontierโ (Nov 18).
Researchers are running 68% more experiments per person than the 2025 average.
Sounds like science is moving faster.
But OpenAI acknowledges that more experiments donโt necessarily mean more discoveries.
AI can help us test more ideas. Can it help us find better ones?
2
1
133
How do you fit almost six working days into one?
OpenAIโs new report, โResearch Acceleration: The Jagged Frontierโ (Nov 18), has an interesting answer.
Researchers now average 5.7 eight-hour blocks of AI agent activity per day.
Not six days of completed work, but machines working in parallel.
Perhaps the next productivity revolution is about doing more things at once.
1
1
104
This might be my favourite finding.
OpenAIโs new report, โResearch Acceleration: The Jagged Frontierโ (Nov 18), reveals that researchers still credited humans with the key safety and post-training ideas behind their latest models.
AI helped turn those ideas into working systems. Some other ideas emerged through collaboration.
Maybe knowing what to build will become more valuable than knowing how to build it.
๐ค Made with AI
1
1
129
The $23,000-a-Day Researcher Is Here.
OpenAI dropped a fascinating research report today, November 18, titled โResearch Acceleration: The Jagged Frontier.โ
One finding jumped out.
Its top 10% of researchers are using AI agents at a rate equivalent to $23,000 a day at public API prices.
Thatโs not an actual bill. And that usage has more than doubled in just one month.
Weโre moving from researchers working with AI assistants to researchers managing entire teams of AI agents.
But hereโs the bigger question:
does 100 times more AI work mean 100 times more discoveries?
1
143
OpenAI asked 101 researchers what theyโd give up to keep frontier coding agents for three months.
The median answer?
40% of team headcount.
Hypothetical, of course. But when people start valuing AI in terms of colleagues rather than tokens, something fundamental is shifting.
1
34
I donโt need another AI news feed.
I want to connect with people who build, break, test and question things.
If thatโs you, letโs #connect.
21
8
426
Gul Saeed Khattak retweeted
6 self-hosted alternatives to Zapier, IFTTT and Airflow
Zapier's pricing goes up with every automation you add.
At first it's a few dollars to send a Slack message when a form comes in. A year later you're running 40 automations and paying more than for your actual software.
GitHub list of free software you can run yourself, has a whole section of automation tools you can host for free:
โก Activepieces: no-code automation like Zapier. Connect apps, set a trigger, done
๐ค Huginn: build agents that watch the web and act for you, years before "AI agents" was a buzzword
๐ changedetection.io: get alerted when any web page changes, like a price, a job listing or a "back in stock" notice
๐ท๏ธ Discount Bandit: tracks prices and stock across Amazon, eBay, Walmart and more
๐ Cronicle: schedule and run tasks from a web dashboard
๐ช Apache Airflow and Kestra: heavy-duty workflow tools for data teams
For non-technical people:
These are little robots that do repetitive computer chores for you while you sleep, without a monthly bill.
For technical people:
Activepieces is the closest drop-in for Zapier. changedetection.io is the hidden gem: point it at any page and it emails you the moment something changes.
The honest catch:
when an automation breaks at 3am, there's no support team. It's you.
But if you're already paying for automation by the task, you'll want to see this.
Links in reply ๐
Take your smart home back from the cloud
Your smart lights stop working when the internet goes down.
Think about that. The bulb is in your house. The switch is in your house. But the decision to turn it on happens on a server in another country.
And when the company shuts down that server, your "smart" home becomes a dumb one overnight.
awesome-selfhosted, the GitHub list of free software you can run yourself, has a fix:
๐ Home Assistant: one app to control lights, heating, locks and sensors from almost any brand, running locally
๐ Tasmota: free firmware that cuts cheap smart plugs off the cloud and gives you full local control
๐งฉ Node-RED: drag-and-drop flows to wire devices and services together
๐ evcc: charges your electric car when your solar panels are producing the most
โ๏ธ SOLECTRUS: a solar dashboard showing what you produce, use and save
For non-technical people:
It's a smart home that keeps working when the internet doesn't, and doesn't tell anyone when you got home.
For technical people:
Home Assistant is the centre of the self-hosted smart home world. Tasmota on ESP-based plugs, plus MQTT, is the classic way to get devices off the cloud.
The honest catch: setup takes an afternoon, and some brands make local control deliberately hard.
But once it's running, your house answers to you, not to a server.
Links in Repo ๐๐
3
5
255
๐ Home Assistant: github.com/home-assistant/coโฆ
๐ Tasmota: github.com/arendst/Tasmota
๐งฉ Node-RED: github.com/node-red/node-red
๐ evcc: github.com/evcc-io/evcc
โ๏ธ SOLECTRUS: github.com/solectrus/solectrโฆ
1
28
6 self-hosted alternatives to Zapier, IFTTT and Airflow
Zapier's pricing goes up with every automation you add.
At first it's a few dollars to send a Slack message when a form comes in. A year later you're running 40 automations and paying more than for your actual software.
GitHub list of free software you can run yourself, has a whole section of automation tools you can host for free:
โก Activepieces: no-code automation like Zapier. Connect apps, set a trigger, done
๐ค Huginn: build agents that watch the web and act for you, years before "AI agents" was a buzzword
๐ changedetection.io: get alerted when any web page changes, like a price, a job listing or a "back in stock" notice
๐ท๏ธ Discount Bandit: tracks prices and stock across Amazon, eBay, Walmart and more
๐ Cronicle: schedule and run tasks from a web dashboard
๐ช Apache Airflow and Kestra: heavy-duty workflow tools for data teams
For non-technical people:
These are little robots that do repetitive computer chores for you while you sleep, without a monthly bill.
For technical people:
Activepieces is the closest drop-in for Zapier. changedetection.io is the hidden gem: point it at any page and it emails you the moment something changes.
The honest catch:
when an automation breaks at 3am, there's no support team. It's you.
But if you're already paying for automation by the task, you'll want to see this.
Links in reply ๐
โก Activepieces: github.com/activepieces/actiโฆ
๐ค Huginn: github.com/huginn/huginn
๐ changedetection.io: github.com/dgtlmoon/changedeโฆ
๐ท๏ธ Discount Bandit: github.com/Cybrarist/Discounโฆ
๐ Cronicle: github.com/jhuckaby/Cronicle
๐ช Kestra: github.com/kestra-io/kestra
18
Gul Saeed Khattak retweeted
A Dead Manโs Switch You Run Yourself
Think about what lives only in your head. The password to the family photos. Where the insurance papers are. The thing you never quite said to your sister.
If you were gone tomorrow, ๐๐ต๐ผ ๐๐ผ๐๐น๐ฑ ๐ธ๐ป๐ผ๐ ๐ฎ๐ป๐ ๐ผ๐ณ ๐ถ๐?
LastSignal is a small open-source app built for that question. The README calls it a self-hosted, email-first "๐ฅ๐ฆ๐ข๐ฅ ๐ฎ๐ข๐ฏ'๐ด ๐ด๐ธ๐ช๐ต๐ค๐ฉ."
You write messages for the people you care about. Every ๐ฏ๐ฌ ๐ฑ๐ฎ๐๐ it emails you a check-in, and you click to say you're fine.
Miss one and it sends ๐ฏ ๐ฟ๐ฒ๐บ๐ถ๐ป๐ฑ๐ฒ๐ฟ๐, ๐ฎ ๐๐ฒ๐ฒ๐ธ ๐ฎ๐ฝ๐ฎ๐ฟ๐. With the last one, it can ask a trusted contact whether you're really unreachable. If they say you're OK, delivery pauses. If nobody answers, your messages go out.
Every timing is yours to change. You can even make one person wait extra days before they can open theirs.
๐ฃ๐ฟ๐ถ๐๐ฎ๐ฐ๐ ๐ถ๐ ๐๐ต๐ฒ๐ฟ๐ฒ ๐ถ๐ ๐ด๐ฒ๐๐ ๐๐ฒ๐ฟ๐ถ๐ผ๐๐. Messages are encrypted before they leave your browser, so the server only ever stores scrambled text. Per the README, even the person running it can't read them.
And it's yours. ๐ก๐ผ ๐ต๐ผ๐๐๐ฒ๐ฑ ๐๐ฒ๐ฟ๐๐ถ๐ผ๐ป, ๐ป๐ผ ๐๐๐ฏ๐๐ฐ๐ฟ๐ถ๐ฝ๐๐ถ๐ผ๐ป. The authors say plainly that anyone selling LastSignal as a service isn't them. You run it with Docker on your own server, even a Raspberry Pi 4 or 5. It's MIT licensed, built with Rails 8 and SQLite.
Google has had something similar for years (Inactive Account Manager), but that hands over your Google data, held by Google.
LastSignal sends words you wrote, from a machine you control.
Two honest caveats:
The README warns that if someone steals the database, they can try to guess your recipients' passphrases offline with no limit, so ๐๐ฒ๐ฎ๐ธ ๐ฝ๐ฎ๐๐๐ฝ๐ต๐ฟ๐ฎ๐๐ฒ๐ ๐ฎ๐ฟ๐ฒ ๐๐ต๐ฒ ๐ฟ๐ฒ๐ฎ๐น ๐ฟ๐ถ๐๐ธ. And it only works if your server and email provider are still running when the day comes.
๐๐ฐ๐ด๐ต ๐ฐ๐ง ๐ถ๐ด ๐ฑ๐ญ๐ข๐ฏ ๐ง๐ฐ๐ณ ๐ฐ๐ถ๐ณ ๐ฎ๐ฐ๐ฏ๐ฆ๐บ. ๐๐ฆ๐ณ๐บ ๐ง๐ฆ๐ธ ๐ฑ๐ญ๐ข๐ฏ ๐ง๐ฐ๐ณ ๐ต๐ฉ๐ฆ ๐ญ๐ข๐ด๐ต ๐ต๐ฉ๐ช๐ฏ๐จ ๐ธ๐ฆ ๐ธ๐ข๐ฏ๐ต ๐ต๐ฐ ๐ด๐ข๐บ.
Links to repo: github.com/giovantenne/lastsโฆ












