Digital3D

Claude Code: Meta acquires Conductor, how to run several agents in parallel without it?

Meta acquires Conductor: read only on November 13, deletion on the 20th. How to export, and run several Claude Code instances in parallel with git...

Read full article

AI: at Anthropic, AI is making research move twice as fast, should we be happy about it?

The METR watchdog measures that Anthropic is improving its AIs twice as fast thanks to AI. What that changes for you, and what was redacted from the...

Read full article

Security: Google's AI has found two vulnerabilities in old RSA encryption, are you affected?

DeepMind's AI agents found two vulnerabilities in 1024-bit RSA. Your bank isn't affected, but your old devices and emails might be. Check your key.

Read full article

Bitcoin: how to trap a thief equipped with a quantum computer?

Cory Fields proposes forcing every quantum thief to put down a deposit that the real owner can confiscate. How it works, and its limits.

Read full article

Quantum: an AI solved a problem that has been open since 2008, what does that change?

An Anthropic model proved a quantum conjecture open since 2008 in three days. What the proof says, its caveats, and what it changes for you.

Read full article

ChatGPT can now make animated GIFs: I tested OpenAI's new image model

ChatGPT creates animated GIFs with Images 2.6. My test against 2.5: same quality to the eye, variable speed, and an image three times more expensive.

Read full article

Security: an OpenAI AI found 948 flaws in Linux, is your phone affected?

OpenAI's AI found 948 flaws in the core of Linux in a single pass. What it changes for your Android, your router, and why it's good news.

Read full article

Security: is one poisoned document in a thousand enough to turn an AI agent into a gateway for hackers?

Berkeley and Anthropic hid a backdoor in AIs via 0.1% of their training texts. The trap, its limits, and the right precautions.

Read full article

AI: is what you think you know about ChatGPT already outdated?

Stanford retested 50 findings about AIs with 20 recent models: 44% are outdated. What has changed, what remains, and the sycophancy that persists.

Read full article