Open-weight models, open tools and datasets, and licensing. 12 articles in the last 30 days.
This is a linkpost for the write-up on my site ; the full body is below, and the code, decisions ledger and devlog are in the repo . Base-model z-score calibration of the J-lens helps elicit hidden secret words from Cywiński et al.'s taboo organisms: 0.805 lea
The Oversight Board is asking Meta to prioritize AI content when it's reported.
Planned 2029 debut could make this Apple’s first enterprise server in decades.
I used an open-source map of a fruit fly’s brain to vibe code a website called PitchFly. Its headline suggestions were delightfully bananas.
Meta will soon unveil new camera-free smart glasses, according to The Information.
Ars previewed Mozilla’s report on how cheap open models caught up on capability.
Downloading a model is increasingly easy. Understanding how it was made, or changing a system at its root, is another matter
I need to detect AI-generated vs human-written text in a Laravel application. Most AI detectors I've tried (including SuperAnnotate's Hugging Face model) produce too many false positives: scoring clearly human-written text as 99% AI. What I've Tried: SuperAnno
Attack Success Rate is usually reported as (successful attacks) / (attempted attacks). For a fixed attack set that is unambiguous. For a search-based attack, it is not, and I cannot find a treatment of this. If my attack is a search over instruction perturbati
Using GPT-6 Astra, Playco built three themed game prototypes from one grey box foundation and reported 50% fewer manual fixes than with the previous model.
OpenAI’s new report explores how students and educators use ChatGPT to make learning more continuous, with support that extends beyond the classroom.
Every byte moved has a cost. As model checkpoints grow to hundreds of gigabytes or even a terabyte, that cost adds up quickly. To make things even worse, moving...
