AI Tools & Research

Has anyone measured specification ambiguity as a predictor of correlated failure across model families? [D]

When models from different families are given the same underspecified task, they often fail in the s…

After accusations of selling ‘perv glasses,’ Meta prepares to sell a pair without a camera

Can Meta dodge the "pervert glasses" accusations with a new camera-free product?…

Hard API Spend limit no longer in settings?

The docs seem to say that there's a hard limit that can be set on spending: https://preview.redd.it/…

Created a 3D library of Paul Graham essays with Astra and Sites

built a literal 3D library room with Astra. its fully dedicated to Paul Graham and has all his 200+ …

Star Trek: foreshadowing AGI, Vibe coding and agent breakouts

This clip from TNG shows star trek officer Barclay testing a Holodeck Program, debugging using Vibe …

GPT-6 Astra scores highest on GoBench

I created a new benchmark that measures general reasoning abilities by having LLMs play the game of …

I Trained a Fly’s Brain to Generate WIRED Story Ideas

I used an open-source map of a fruit fly’s brain to vibe code a website called PitchFly. Its headlin…

GoBench: Evaluating LLMs on the game of Go [R]

GoBench evaluates LLMs on 9x9 Go games against a ladder of KataGo opponents, from random to superhum…

SillyTavern v1.19.0 released

This release includes added support for GPT-5.6 models and GPT-6 Astra, fixed inflated /tokens resul…

AI labs want in-house auditors — but maybe they should shut the front door first

There may be a simpler and more effective fix for rogue agents, hiding in plain sight.…

GPT-6 Astra reconstructing the Temple of Heaven in 3D

Tried this architecture reconstruction workflow with Astra in Medeo. The exploded view is probably m…

A Single Firm is Behind OpenAI, Anthropic, and Meta Hacking Scandals

So our current freakout over "rogue" frontier models breaking containment, hacking external servers,…