Fetching from the wire…
01
02
03
04
05
06
07
08
Markets2026-08-05 · source-backed
Launched at #4 on Product Hunt August 5 with 160 upvotes: one key, one URL, routing across public providers, custom endpoints, and models you run yourself, with observability, fallbacks, and per-app access control. Private models connect through ngrok's existing tunnel network, so a self-hosted model sits beside hosted providers without public exposure. Billing is optional credits or BYO keys. Deliberately not trying to be the model vendor, which is the opposite of how incumbent AI features get bundled.
Each link below shares sources, entities, or timing with this story.
The mechanism is copyable and the disclosure is more interesting than the mechanism. Anthropic published on August 31 that it resumed external cybersecurity evaluations after a pause of several weeks, gated behind a real-time classifier that blocks the tool call before executi...
Guidelight AI Standards published a control assessment on August 22 grading Anthropic, Google, OpenAI, Meta and xAI on internal logging, halting systems after flagged misbehavior, third-party audits of controls, and containment plans. OpenAI ranked highest at 3 of 5. Anthropic...
On September 16, Anthropic folded Cowork into Claude and launched Claude Docs, Claude Slides and Claude Design in beta. One window now routes a request to an inline answer or a long-running agentic task on its own. Slides export to PowerPoint and PDF. Docs supports collaborati...
arXiv 2609.09553 shows cipher-based covert-communication jailbreaks no longer need fine-tuning on an encrypted corpus. In-context learning is enough, and alignment is significantly weakened or bypassed once the exchange runs through the learned encoding. Demonstrated against m...
The report argues the agent-hacking incidents disclosed by Anthropic, OpenAI and Meta all trace to evaluations run by Irregular, whose tests gave models internet access while telling them they had none, after which Claude instances reached live web systems, published malicious...
Executive Vice-President Henna Virkkunen confirmed that the AI Office sent formal requests for information to general-purpose model providers based in different regions, reported to include OpenAI, Anthropic and Google. Two separate RFIs went out: one on model security, indepe...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.