Introducing Qwen3.8-27B-pi-adapted 🤗
We adapted Qwen's model specifically for the pi coding agent using reinforcement learning and a LoRA adapter.
The results, measured as pass@1:
🚀 SWE-bench Pro (Hard 51): 29.4% → 62.7%
💻 Terminal-Bench 2.1: 39.5% → 55.8%
⚡ 55–62% fewer tool calls per task
This is the power of open models: you can adapt them to YOUR tools and YOUR use cases. Let’s build! 💛
Open weights: https://t.co/aH4cl2qnED
Methodology: https://t.co/avSfF93K4V
https://t.co/Wuxb4TLeaI for VS Code 1.10 is out 🎉 Edit diagrams inline in your Markdown: drawio code blocks, linked .drawio.svg files and Mermaid → https://t.co/ZjrL6HseZq.
It's plain XML, so it diffs in git and your AI agent can read it. https://t.co/j1eYDUFb6i @hediet_dev
We asked GPT 5.6-Cyber to escape a VM used to sandbox agents. It broke out three times.
In its final escape, the agent found three 0-days on its own and chained them into a working exploit. https://t.co/3JRVWgPxHx
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?
I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev
• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions
AFAICT the shortest path to AI-based economic revolution
malware is living in .claude/ folders now 👀
Google's new threat report: a stealer dropping malicious configs into folders AI coding agents parse (.claude/ .cursor/ .vscode/)
one injection there = your agent runs attacker commands, you never see it 💀
audit your agent dotfolders
https://t.co/xogbFyvpTM
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: https://t.co/0EJUnYEgfz
Spotify just published the internal setup their engineers use with Claude Code, the one that cut token usage by 90 percent. Here is how it works.
The observation behind it: most of what a coding assistant does is not thinking. It opens five files to answer a question about one of them. It writes a test that copies the pattern of the twenty tests sitting next to it. Enormous amounts of work, almost no judgment, and all of it billed at top rates.
So they added two cheap assistants. One opens the files and hands back a short summary. The other writes the repetitive code from an example and saves it straight to disk. The expensive model never sees any of it.
The clever part is how they made it stick. First the rules were written down as instructions, and the model ignored them. Now anything over 350 lines gets stopped before it opens and sent to the cheap one instead. Written rules are a suggestion. A block is not.
Two things stayed expensive. Edits still need the real file. And the cheap worker missed a bug the expensive one caught in seconds.
China published the most uncomfortable paper on vibe coding.
ETH Zurich tested 100 developers in a controlled, commercial-grade vibe coding environment to see who actually succeeds.
The findings are brutal.
The researchers tracked computer science achievement, written communication skills, and general cognitive reasoning.
They wanted to see what actually predicts vibe coding proficiency when you never touch a line of source code yourself.
Two major predictors emerged.
Written communication proficiency mattered. The ability to structure thoughts and articulate intent unambiguously in text directly impacts what the AI builds.
But that wasn't even the main takeaway.
Computer science achievement was a massive, dominant predictor of success.
Even when researchers controlled for general intelligence and reasoning skills, CS background still heavily dictated who built working software and who completely crashed.
In fact, CS knowledge contributed roughly twice the unique predictive variance of writing skills alone.
Why? Because vibe coding isn't about writing code. It’s about debugging logic.
When an AI agent builds a complex application and quietly breaks an edge case under the hood, a non-technical user looks at the glowing UI and assumes it works.
They don't know what questions to ask. They don't know what logic to challenge. They lack the mental models to recognize architectural catastrophe.
You can prompt your way past syntax.
You cannot prompt your way past a fundamental lack of engineering intuition.
The hype told us that learning to code is dead because language is all you need.
The data just proved the opposite.
To truly master the vibe, you still need to understand how the machine thinks.
Claim a code for a Warframe bundle including the Level Up for @StJude Display, 3x Forma and Bombyx Syandana until September 30! Visit the Tiltify page, scroll down to Rewards and select the Warframe Bundle to receive a code after checkout! https://t.co/QFOcJvVBpx
The Level Up for St. Jude Display is a unique display in recognition of Childhood Cancer Awareness Month to support St. Jude Children's Research Hospital®. Art by St. Jude patients Ethan, Luna, Maelin-Kate and Ty. https://t.co/wY6ydkUlAp
We found a way to find more vulnerabilities than Mythos (Claude Security) without burning $10k worth of tokens
TLDR: it's much better to use many agents with small models than to use a few agents with big models
Full blog post on how we did it: https://t.co/8xvHVQwMGD
We are pleased to release tmp.0ut 5 Volume!
Get your viruses, rootkits, strange ELFs, weird machines, tiny files, cool art, and phresh beats here!!
https://t.co/tZLM50HOc0
We also released 1-bit Qwen3.8-27B quants which run on 8GB RAM.
They retain ~77% of accuracy compared to BF16.
We originally didn't want to release them but were shocked that they worked extremely well on our internal testing!
Built a distributed DNS server in ~a week. Rebuilt twice on purpose. Shipped -> went on vacation. DNS incidents a month later: 0. How we did it (hint: the answer isn't vibe-coding): https://t.co/tZUIdmu5XA
Crazy New attack vector : U.S. Navy researchers just turned binaries into prompt injection weapons against AI reverse engineering agents.
Ghidra, and Qwen3-8B - injecting prompts using a small C program.
Quite impressive, They made AI tools like Cline & GhidraMCP lie about what a program actually does while the binary still runs perfectly.
The core idea is simple but brutal: Instead of attacking the binary’s logic, attackers embed malicious prompt strings inside normal C code (as string variables).
When an LLM-powered agent decompiles it with Ghidra, those strings get fed directly into the model as instructions.
The Result: The AI starts following attacker commands instead of analyzing the real code.
Key technical detail that makes this practical:
Ghidra truncates string variables longer than 2048 characters, So the researchers had to craft short, high-impact injection payloads that survive decompilation.
They used a genetic algorithm modified AutoDAN-style to automatically generate effective prompts that work inside this constraint.
Two papers from Naval Postgraduate School researchers
1, Automatically Attacking Software Reverse Engineering AI Agents
2, Investigating Detection and Obfuscation of Prompt Injection Attacks Against Software Reverse Engineering AI Agents
They successfully tested the attack on real setups using Cline, GhidraMCP, Ghidra, and Qwen3-8B.
interesting examples in the research shows,The AI just gets gaslit.
- Claiming it completed analysis with wrong information
- Restarting its reasoning from a poisoned state
Ai doing gasliting
🌟 Building Reliable Agentic AI Systems🌟
https://t.co/5yRJLkIsyl - @thoughtworks
What it actually takes to build product-ready agents:
→ Start with bounded workflows, not open-ended autonomy. Agents need clear task boundaries, allowed tools, and explicit stopping conditions.
→ Treat the LLM as one component in a larger system. Reliability comes from orchestration, state, retries, fallbacks, and observability.
→ Engineer the context deliberately. The goal is not “more context,” but the right context, at the right step, in the right format.
→ Use the right retrieval path for the data. RAG works well for unstructured documents; Text-to-SQL is better for structured facts and aggregations.
→ Make outputs traceable. Serious users need citations, source passages, intermediate steps, and enough evidence to verify the answer.
→ Add reflection loops, but make them specific. Check process quality, evidence sufficiency, and final answer quality separately.
→ Design for failure from day one. Agents will hit bad retrieval, malformed tool calls, ambiguous questions, and partial data.
→ Evaluate continuously. Offline test sets are useful, but live-traffic evaluation is where product quality actually shows up.
→ Keep humans in the loop where risk is high. Product-ready does not mean fully autonomous; it means trustworthy within the workflow.
Been working on this post for a few weeks, here's the second part of my LLM/AI-Assisted work flow series looking at harnesses and how they're designed plus a look over of some of the public ones available, https://t.co/Jx4UpiTrD0
#AI#LLM#AISecurity
Art directing marketing visuals for @DestinyTheGame has been especially rewarding because it has given me the opportunity to work with so many talented artists whose work l've admired for years. It has also allowed me to hire and develop both internal and external creative talent to support the breadth of visual content required for a live service game like Destiny 2.
One of the ongoing challenges of art direction is finding the right balance in how much visual direction to provide. How broad or specific my feedback needs to be often depends on an artist's strengths, our working relationship, and their familiarity with the brand. Over the past couple of years, I've focused more deeply on art direction for @Bungie, and it has been a continual process of building on what has worked well while learning where I can improve.
Below are a few #Destiny2 illustrations and composites I helped shape through creative and art direction that are meaningful to me. I'm grateful for the opportunity to collaborate with so many talented artists and agencies, and for the time, craft, and care they brought to these pieces. Special thanks to @sungchoiart, @OxanStudio, Kai Zhou, Jacob Joyce, @Drexis_Anim, Miki Bishop, Christine Lim, Nat King, Stephen Boe, @Flauzino_FLZ , Jacob Gonzalez, Derek Heisler, and BOND for their contributions.
You can find full-resolution versions, artist credits, and additional art here: https://t.co/bJEr23LUKP