Exclusive | The Sudden Fall of OpenAI’s Most Hyped Product Since ChatGPT - WSJ
Exclusive | The Sudden Fall of OpenAI’s Most Hyped Product Since ChatGPT WSJ
Could not retrieve the full article text.
Read on Google News: OpenAI →Sign in to highlight and annotate this article
Conversation starters
Daily AI Digest
Get the top 5 AI stories delivered to your inbox every morning.
More about
productchatgpt
FairyClaw: An Open-Source Async Agent Runtime for Long-Running, Event-Driven Server Deployments
Hello everyone! We’re excited to share our new open-source project, FairyClaw . No more bloated SKILL.md files eating your context window. Structured code replaces prose — cheaper and more accurate. Meet FairyClaw — an open-source async agent runtime for long-running server deployments: Restructured Skill paradigm : instead of verbose markdown instructions flooding the context, Skills are declared as structured steps in manifest.json — deterministic playbooks that cut token usage while making tool selection and execution order semantically precise Single-step Planner : one inference per wakeup, event-driven continuation — no monolithic loops, tasks resume naturally after interruption Capability Groups : tools, skills, and hooks declared as pluggable units, routed semantically — structured
The Claude Code Leak Changed the Threat Model. Here's How to Defend Your AI Agents.
IntentGuard — a policy enforcement layer for MCP tool calls and AI coding agents The Leak That Rewrote the Attacker's Playbook On March 31, 2026, 512,000 lines of Claude Code source were accidentally published via an npm source map. Within hours the code was mirrored across GitHub. What was already extractable from the minified bundle became instantly readable : the compaction pipeline, every bash-security regex, the permission short-circuit logic, and the exact MCP interface contract. The leak didn't create new vulnerability classes — it collapsed the cost of exploiting them . Attackers no longer need to brute-force prompt injections or reverse-engineer shell validators. They can read the code, study the gaps, and craft payloads that a cooperative model will execute and a reasonable devel
Oil climbs anew on mixed signals about Iran war's future
Brent crude oil climbed more than 1% to above $110 per barrel when markets opened Sunday amid mixed signals about the Iran war that's creating unprecedented disruption to global energy flows. Why it matters: President Trump is signaling major escalation, but also told Axios' Barak Ravid that the U.S. is in "deep negotiations" with Iran. Trump is threatening to bomb Iran's power plants and bridges starting Tuesday if the regime doesn't open the Strait of Hormuz. The big picture: Markets are also responding to crosscurrents about the Strait of Hormuz and regional infrastructure. Iranian officials say they will exempt Iraq from restrictions and a tanker carrying the nations' crude has reportedly transited the waterway. But specifics and conditions for Iraqi crude remain unclear, per Bloomberg
Knowledge Map
Connected Articles — Knowledge Graph
This article is connected to other articles through shared AI topics and tags.
More in Models
Show HN: Gemma Gem – AI model embedded in a browser – no API keys, no cloud
Gemma Gem is a Chrome extension that loads Google's Gemma 4 (2B) through WebGPU in an offscreen document and gives it tools to interact with any webpage: read content, take screenshots, click elements, type text, scroll, and run JavaScript. You get a small chat overlay on every page. Ask it about the page and it (usually) figures out which tools to call. It has a thinking mode that shows chain-of-thought reasoning as it works. It's a 2B model in a browser. It works for simple page questions and running JavaScript, but multi-step tool chains are unreliable and it sometimes ignores its tools entirely. The agent loop has zero external dependencies and can be extracted as a standalone library if anyone wants to experiment with it. Comments URL: https://news.ycombinator.com/item?id=47655367 Poi
![[R] Reference model free behavioral discovery of AudiBench model organisms via Probe-Mediated Adaptive Auditing](https://d2xsxph8kpxj0f.cloudfront.net/310419663032563854/konzwo8nGf8Z4uZsMefwMr/default-img-graph-nodes-a2pnJLpyKmDnxKWLd5BEAb.webp)
[R] Reference model free behavioral discovery of AudiBench model organisms via Probe-Mediated Adaptive Auditing
Anthropic's AuditBench - 56 Llama 3.3 70B models with planted hidden behaviors - their best agent detects the behaviros 10-13% of the time (42% with a super-agent aggregating many parallel runs). a central finding is the "tool-to-agent gap" - white-box interpretability tools that work in standalone evaluation fail to help the agent in practice. most auditing work uses the base model as a reference to compare against. i wanted to know if you can detect these modifications blind - no reference model, no training data, just the target model itself. maybe you can? and the method is embarrassingly simple. LoRA fine-tuning tends to modify later layers more than earlier ones. so i train a Ridge regression from early-layer activations (~L12) to late-layer activations (~L60) and look at the residua
Discussion
Sign in to join the discussion
No comments yet — be the first to share your thoughts!