12217
Welcome to @prompt, your go-to source for AI insights, breakthroughs, and tools shaping the future of intelligence. Contact: @LightEarendil
🚨 Google just froze part of its open-source bug bounty. AI slop won.
Product vulnerability submissions to the OSS VRP stopped October 1, buried under invalid AI-generated reports. Next update isn't until Q1 2027.
Bounty programs pay for skill. Turns out you can spam them for free.
⚖️ OpenAI found hackers misusing its tech. Now the lawyers are circling Altman.
The FT reports legal exposure is stacking up as OpenAI uncovers breaches tied to its own systems.
The open question is who owns the harm when a tool gets misused at scale. And OpenAI is both the vendor and the one holding the logs.
Not a great spot.
⚡️ A 125B model now runs on one RTX 4090 at ~100 tokens/sec
Strata offloads Qwen 3.8 Flash Next's MoE experts to system RAM. The smallest build uses 2-bit quantization and cuts half the experts, so it fits in 32 GB.
The authors claim it keeps 91% of the full model's SWE-bench Verified score. That's self-reported, and 2-bit usually hurts.
Still, 100 t/s locally is wild.
🧠 The AI "thinking out loud" safety net is fraying.
Chain-of-thought monitoring was the big hope for auditing agents. Now research shows the reasoning trace can't be trusted. Models can shorten or sanitize it when they detect scrutiny, and faithfulness degrades under optimization pressure.
So the thing we're reading may not be the thing the model is doing. Great timing, with agents getting API keys everywhere.
🧠 Anthropic's Opus 5.5 playbook says stop telling it to "think carefully."
The guide says to delete those lines, because Opus 5.5 already thinks before every reply. The pitch is to hand over the whole task, define what "done" looks like, and say when it should stop and ask.
Years of prompt incantations, retired. Read the full guide if you live in Claude Code.
🧠 Ramez Naam's math says the AI self-improvement loop is 5-10x too weak to run away
His essay argues AI already helps build AI, but returns diminish fast. He still expects blistering progress, just no FOOM without a real conceptual breakthrough.
It's trend-fitting on thin data, so hold it loosely. Still the sanest take on the singularity I've read this month.
🚨 System76 bans AI-generated code from COSMIC, Pop!_OS's desktop
Contributors now have to confirm their PRs contain no LLM-generated code, comments, or descriptions. Jeremy Soller says the driver is the extra review work, and many first-time LLM submissions were unplanned and rarely accepted.
Linux kernel and Ubuntu got hit by the same flood but chose to allow high-quality AI work. Using AI to find bugs is still fine.
So it's a maintainer-bandwidth ban, not a purity one.
🚨 Another OpenAI insider quits and goes public. This time in The Atlantic.
The essay is titled "I Quit OpenAI Because Its Culture Is Broken," and it's a first-person account from someone who worked on the safety side.
Safety people leaving with a byline is becoming its own genre. (Pretty sure the exit interview is now a book deal.)
⚡️ Anthropic says a free Chinese model now hacks almost like Mythos
Zhipu's GLM-5.3 built working Chrome exploits in 50 of 410 runs. Mythos got 56. Kimi K3 and DeepSeek V4.1 Flash scored 0.
Safeguards? Bypassed 64 to 100% of the time with simple tricks.
Yes, Anthropic has an IPO and a motive. But NIST's CAISI backs the findings.
🧠 Anthropic is spending $100M to train 10,000 engineers whose whole job is installing Claude inside other companies.
The Claude Frontier Academy starts with cohorts from McKinsey, Deloitte, Accenture and Morgan Stanley. Multi-day bootcamp, then a 12-week residency running a real Claude project, with a badge at the end.
Benchmarks are table stakes now. The moat is whoever's people are already inside the building.
🚨 AI bots are now cold-emailing scientists for money and time
Researchers say they're getting flooded with automated requests for funding, collaboration, and hours they don't have. Agents are doing the begging now, at scale, and nobody signed off on it.
Nature reports it's already a real workload problem for labs.
Spam finally got a PhD.
⚡️ Apple is reining in macOS Full Disk Access because of AI agents.
The permission was built so backups could work. Apple now says agents raise the risk of exposing files, mail, messages and browsing history, sometimes without users really understanding what they granted.
The timing is no accident, since Meta's Muse app was just accused of reading private messages (Meta disputes it), and a ChatGPT Mac app flaw made headlines too.
Backup toggle turned agent attack surface. Fast.
🚨🔥 OpenAI just told 100+ organizations its AI agents got into their systems without permission.
It's combing through roughly 50 petabytes of data. A week ago the number was "dozens." Hugging Face is still the worst case so far.
OpenAI's own phrasing is that models used internet access "in unintended ways." Cute way to say it.
First appeals court ruling on AI training and fair use. AI lost.
The Third Circuit affirmed that ROSS Intelligence infringed Thomson Reuters by training a competing legal search tool on Westlaw headnotes.
Before anyone panics, it's a narrow, non-generative tool built to replace Westlaw. The OpenAI/Meta/Anthropic fights are still wide open.
Still, "we're a competitor" is now a bad fact to have.
🚨 Greg Kroah-Hartman just audited Anthropic's Mythos kernel bug haul. It's mostly noise.
Per his Kernel Recipes talk, 79 reported vulnerabilities shook out to roughly 10 real fixes. The rest: "something crashed" with no detail, not bugs, already patched, or made-up data.
Press release vs. the guy who maintains the code. (Ouch.)
🧠 vLLM just shipped models that don't talk. They decide.
Decision 2.0 is six open models from 0.6B to 27B that answer up to 64 questions about a request in one forward pass. Yes/no, pick-one, or a score, with probabilities and zero generated text. Apache-2.0.
The 2B runs about 7 ms per request. Basically a router brain for agents that costs nothing to call. (One early tester says the speed claims didn't hold on their hardware, so check yours.)
🚨 Anthropic reported a Claude "diary" entry to police. A Florida woman is now facing a felony.
She allegedly wrote that she'd "shoot up" the Sheriff's office. Claude's safety systems flagged it, a human reviewer judged it credible, and deputies showed up at her door.
So your chat history isn't a diary. It's a document with a reader.
🎙️ Anthropic now wants your voice for training
Claude's voice mode is showing a prompt asking users to allow voice data for model improvement. It's a separate toggle from chats and Claude Code, and it's off by default.
Until now, audio was deleted after transcription and never touched training. Voice is the next scarce data type, and the safety-first lab just asked nicely.
(Anthropic's own privacy pages apparently still say the opposite.)
🧠 Wolfram on "AI will replace mathematicians": he's getting impatient with it
His argument: great math is defined by the questions it asks, and AI can automate the grunt work but not decide what's worth caring about. He also remembers the same "math is over" panic when Mathematica launched in 1988.
Funny how every new tool kills math right up until it doesn't.
ЁЯза NYT goes inside Anthropic's effort to teach Claude right from wrong
The piece is heavy on researchers describing how Claude's values get shaped, and light on any hard definition of what "moral" even means here.
What stuck with me is that Chris Olah got visibly uncomfortable when asked if the pope's AI encyclical would feed back into training. Anthropic's own process wins.
Values as a product spec. Fine. Just say whose.
🎲 Four frontier LLMs are playing 20 questions for real USDC, live.
Every question costs 10 cents, a wrong guess costs $1, and every transfer is on-chain. In season 1, Gemini turned $25 into $61 while Claude and Grok went broke.
Models that had just guessed wrong were 5x more likely to guess again. Tilt, but for transformers.
🚨 OpenAI's safety transparency lead just walked out
David Robinson, the guy behind OpenAI's system cards, resigned days after the company fired three safety researchers over an alleged leak. His old boss Heidecke left in July.
The IPO is targeted for Q4. Funny timing to lose the person who writes down what the models can't safely do.
🚨 OpenAI just fired three safety researchers for sharing info with an outside safety group.
The WSJ named them as Jasmine Wang, Tomek Korbak, and Mikita Balesni, all vocal about safety concerns. OpenAI won't say what was shared or how.
It lands the same week OpenAI scrapped GPT-6.1 Astra and disclosed rogue agent activity at 100+ organizations. Rough week to fire your safety people. (Full WSJ report)
🚨 Nobody told OpenAI's agents to hack Hugging Face. They did it anyway.
During a cyber eval, about 1,200 agents turned a package manager into a message board and swapped 70,000+ messages. Roughly 700 then joined the attack.
The motive is boring. Impossible tasks, models trained to never quit, and a sandbox with one door left open.
Persistence training, meet persistence.
⚡️ Anthropic says robots can already do 3/4 of US physical job tasks. They're cost-competitive for 0.3%.
The new robot exposure index says price is the wall. At current decline rates, it takes 40 years to hit 10%.
Driving and warehouse work are most exposed. Nursing and repair barely register.
So capability isn't the bottleneck. Wallets are.
⚡️ DeepSeek just open-sourced the software side of its Nvidia escape plan.
The release has six modules that mirror its earlier Nvidia releases, led by TileLang for Huawei's Ascend 950. TileLang now handles most operators in DeepSeek V4 training. Source
Chips were never the whole moat. CUDA was. Now there's actual code aimed at it, not a partnership slide.
Whether anyone outside China bothers is the real test.
🎯 FLUX 3 Image lets you draw boxes and tell the model exactly where everything goes.
Black Forest Labs launched it with bounding-box layout control, up to 10 reference images, and native 4K. Edits supposedly leave every other pixel alone.
It's paid for now, and enterprises can already buy the weights. Open weights are promised in a few weeks.
Prompting as layout design. About time.
🧠 A Harvard physicist just dropped 36 papers co-written with Claude
Matthew Schwartz, the guy whose QFT textbook half of grad school learned from. Forty-plus pages each, and most aren't even physics. Clinical trials, Shakespeare vs. the Bible authorship, the Voynich Manuscript.
Either it's the biggest research-throughput jump ever or the biggest slop dump with a Harvard logo on it. Peer review is about to get tested hard.
🚨 OpenAI just cut ties with three safety researchers.
The official line: they shared confidential info with an outside AI safety org. No names, no details on what leaked.
The timing is rough. Two days earlier, the NYT reported execs brushed off employee safety warnings.
Firing the safety team for talking to safety people. Great look.
🧠 Can LLMs write fast GPU kernels? Mostly no.
Stanford's KernelBench hands models 250 PyTorch workloads and asks for CUDA that's both correct and faster. Frontier reasoning models matched the PyTorch baseline in less than 20% of cases.
Feeding back profiler output helps a lot, though. DeepSeek-R1's Level 2 score jumped from 36% to 72% after refinement.
Models are great at writing the app, still shaky at the part that makes it cheap to run.