Google ships Gemini 3.7 Flash three weeks after the last one
Google says its new entry-level model beat Anthropic and OpenAI across nine benchmarks. The cadence matters more than the scores.
Read More
The self-hosted AI stack: what to run at each layer
You can run a complete AI stack on hardware you control. Whether you should depends on what an hour of maintenance is worth to you.
Read More
AI has cut entry-level hiring 14% while total employment holds
An effect concentrated in one part of the workforce can be invisible in a national average and severe for the people inside it.
Read More
How multimodal AI models actually process images
The mechanism is simpler than the capability suggests, and unusually predictive of where these models fail.
Read More
Mechanistic interpretability explained: reading a model from the inside
We specify a training process, not a design. Nobody writes down what the finished network does, which means nobody starts out knowing.
Read More
US chip export controls reversed twice in five months
The test is now where a company is headquartered, not where the chips are delivered. Nvidia forecasts zero China data-centre revenue.
Read More
The right to be forgotten meets a model that cannot forget
European law gives you the right to have your personal data deleted. A trained model has no mechanism for forgetting anything.
Read More
Ten governments can now test frontier AI, but none can block a launch
AI safety institutes evaluate models on access the labs grant voluntarily. The clearest signal of where they are heading is what they have been renamed.
Read More
33 states now regulate election deepfakes, with five different rules
Thirty-three states now regulate AI-generated political content. Two had their laws struck down, and there is still no federal baseline.
Read More
How to run an LLM locally: what fits on your hardware
The main thing between you and a working local setup is picking a model that fits your memory. Get that wrong and you will conclude local models are useless.
Read More
When to fine-tune an AI model, and when not to bother
Fine-tuning teaches behaviour. It does not teach facts reliably, and confusing those two is where the money goes.
Read More
How AI image generation works, and why prompts fail
Nothing is composed. The image emerges everywhere at once, which is exactly why counting and text placement go wrong.
Read More