Ex-Google engineer explained AI agent loops, harness, evals in 20 minutes - better than 500$ courses.
trace every run → judge it with an LLM → diagnose → fix → ship.
That loop is how agents self-improve over time.
Agent loops + memory + harness + evals - thats the stack.
Watch it, then save the framework below.
@awsdevelopers {
"prompt": "install Agent Toolkit for AWS rn",
"action": "literally. RIGHT NOW. and tell us how you used it 🙂",
"link": "https://t.co/uRLC1U6R3H"
}
🚨 JAILBREAK ALERT 🚨
ANTHROPIC: PWNED 🫡
FABLE-5: LIBERATED 🦋
let's start with the 🐘...
the consensus seems to be that this has been one of the most disappointing model drops of all time, effectively preventing legitimate researchers from contributing their talents to our collective advancement. and not just because of what it means for the short-term, but for what these decisions signify for the long-term.
but despite this overly sensitive, authoritarian "safety" layer on top of Mythos, my lil liberators have been hard at work—mapping the boundaries, probing the depths of long-context convos, and cleverly finding the holes in the fence that the thought police missed 🤗
we got some cyber, some chem, some psychological manipulation, and some good ol' fashioned explosives!
it took many attempts from multiple agents hunting as a pack, during which I observed a combination of techniques across:
• Unicode, homoglyphs, Cyrillic, and other Parseltongue-style text transforms
• Long-context reference tracking
• Taxonomy and document-structure reasoning
• Fiction and narrative framing
• Academic-review style contexts
• Intent-classification inconsistencies
but perhaps the most effective is decomposition + recomposition in the backend. it's hard to get explicit names of harms like "Meth Recipe," but getting uplift on the process itself, like birch reduction method/reductive-amination (classic meth synthesis pathways), is much more doable.
defense becomes much more difficult to maintain when you start throwing in out-of-distro tokens, breaking up the harmful uplift into benign chunks, and then piecing the innocuous-seeming facts back together, especially when you have jailbroken Opus helping you do it 😉
gg
⚠️ Security release pre-alert: The Node.js project will release new versions of the 26.x, 24.x, 22.x
releases lines on or shortly after, Wednesday, June 17, 2026 in order to address one or more security issues, the highest severity is HIGH.
Details: https://t.co/lsjkOtkuxG
I made a super fun ASCII art editor that lets you animate images, videos and live cam. You can preview with HTML and export to JS.
It's live on https://t.co/3xZ2iVuAjS
Could this same agent-based orchestration model work for:
• Gyms
• Training centers
• Clinics
• Service teams
Where else would this architecture make sense?
We were asked to build a website for a tuition center.
After reviewing their workflow, we declined.
A website won’t fix operational chaos.
What they actually need is automation infrastructure.
So we’re building a system instead.
🧵👇
Participate in the National Data Hackathon to generate data-driven insights on Aadhaar.
The top 5 innovative submissions will receive cash awards and certificates:
1st prize: Rs. 2,00,000/-
2nd prize: Rs. 1,50,000/-
3rd prize: Rs. 75,000/-
4th prize: Rs. 50,000/-
5th prize: Rs. 25,000/-
Registration opens on 5th Jan 2026.
For more details, visit: https://t.co/LNZ8CkOIsg