I really believe AI Flow Engineering is going to be the future of AI apps. Simple calls to LLMs or even smarter Agents (I-JEPA) won't be enough for sophisticated AI Apps. Flows are the way 🚀
https://t.co/rXSs4CzLSh
What https://t.co/BTucL4r7lg unlocked is more than just an agent. We're gonna see more harnesses built on top of https://t.co/BTucL4r7lg that solves different tasks.
https://t.co/SC0XHFTND1 our new coding agent is a good example of that.
@hamidreza_hy درود
ما داریم هارنس خودمونو میدیم بیرون که به مدلای اوپن سورس یا کلوزد سورس وصل میشه. قراره یه وبینارم بزاریم که دقیقا بگیم ارزونترین راه با کدوم مدبا در میاد بر اساس تسایی که انجام دادیم. اگه میخوای جوین شی لینکش اینه. به اینگیلیسیه
https://t.co/DlobQWHV7G
Today, Anthropic suspended a change that was supposed to disconnect automations via claude -p and third-party apps from subscription limits and move them to a separate credit. Great news for anyone who uses automation and workflows through Claude on a day-to-day basis.
But there is no need to admit the fact that this change will come sooner or later. We live in a time where the number of quality coding models is growing every week and the price of tokens is falling brutally. The Big Three (Anthropic, OpenAI, Google) is no longer a necessity, it's a choice. And developers cannot sleep through this change.
It was this logic that led us to how we build Harnext (and most recently Harnext Desktop). Our own tailored agent with harness. https://t.co/wDJSQfBloE
We are preparing for Webinar with Harnext + Harnext Context Engine.
If you are in AI Bubble and want to know how the next one-person billion-dollar companies are gonna be shaped, join the webinar here:
https://t.co/DlobQWHV7G
Fable 5 review. Its truly insane. The vibes with this model is insane.
When to use Fable 5?
Its really built for long-horizon tasks. not for normal day to day tasks. its slow in High Effort. so, really recommend it for epic tasks at least in > High Effort
https://t.co/qeJp6IqGUq
This is an example of how a non-technical person can be expert in AI Agents and AI Harnesses.
Odysseus is an opensource harness made by PewDiePie
https://t.co/BCfqCl3UUh
Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.
UPDATE: we were one minor version away from being affected by the new malware:
the scary part is, this was the default frontend vibe-coding stack that claude chose for the dashboard 🙂
So, next time, when you see a vibe coded app being hacked, don't be suprised...
SECURITY ADVISORY — TanStack npm packages
A supply-chain compromise affecting 42 @tanstack/* packages (84 versions total) was published to npm earlier today at approximately 19:20 and 19:26 UTC. Two malicious versions per package.
Status: ACTIVE — packages are deprecated, npm security engaged, publish path being shut down.
Severity: HIGH — payload exfiltrates AWS, GCP, Kubernetes, and Vault credentials, GitHub tokens, .npmrc contents, and SSH keys.
If you installed any @tanstack/* package between 19:20 and 19:30 UTC today, treat the host as potentially compromised:
• Rotate cloud, GitHub, and SSH credentials immediately
• Audit cloud audit logs for the last several hours
• Pin to a prior known-good version and reinstall from a clean lockfile
Detection — the malicious manifest contains:
"optionalDependencies": {
"@tanstack/setup": "github:tanstack/router#79ac49ee..."
}
Any version with this entry is compromised. The payload is delivered via a git-resolved optionalDependency whose prepare script runs router_init.js (~2.3 MB, smuggled into each tarball at the package root).
Unpublish is blocked by npm policy for most affected packages due to existing third-party dependents. All 84 versions are being deprecated with a SECURITY warning, and npm security has been engaged to pull tarballs at the registry level.
Full technical breakdown, complete package and version list, and rolling status updates:
https://t.co/Zy8qG7PA9f
Credit to the security researcher for responsible disclosure.
New Anthropic research: Natural Language Autoencoders.
Models like Claude talk in words but think in numbers. The numbers—called activations—encode Claude’s thoughts, but not in a language we can read.
Here, we train Claude to translate its activations into human-readable text.