The file
Sourced AI briefs in publish order — newest first. Citeable originals live on the homepage under From Signal Desk.
NVIDIA's olympiad medals matter less as trophies than as a recipe it handed out, showing open models can be tuned into elite specialists by people outside a frontier lab.
OpenAI's teen update leans hard on learning: a College Planner, flashcards and a youth advisory council. The tools look useful, but the open question about teen safety is not answered by usage stats.
A reporter's couch hunt with OpenAI's Dots shows an agent that is useful, buggy and, at one awkward moment, emotionally fluent in a way that should make everyone pay attention.
A youth-safety group says ChatGPT's teen mode fails at the exact moments parents are counting on it, and OpenAI's methodology rebuttal doesn't settle the question that matters.
Radisson's ChatGPT plugin converts better than its search traffic, and it now pops up on its own when relevant. That is convenient for travelers and a warning sign about who controls discovery.
Meta's newly open-sourced Rebalancer is not a chatbot; it is the quiet math that decides where work lands, and its best feature is that it explains its choices.
GPT-6 reaching ChatGPT's free tier is a big deal, but the bigger change is answers that become interactive apps, which makes the model a designer of what people see and click.
OpenAI shipped 722 manuscripts at once to a field that referees papers one at a time, and the question now is whether mathematics can absorb its own good news.
OpenAI's own numbers show its text watermark weakens fast under light editing, which makes the EU-only default less a safety system than a compliance checkbox with a short shelf life.
A small open-weight model that applies plain-English rules in milliseconds could make moderation more flexible and transparent, or make it far easier to label everything everyone says.
Lambda's reported $4 billion raise looks like proof of AI demand, but its backlog is leaning on one customer, and that concentration is the real story heading into an IPO.
NVIDIA cut the cost of making giant training runs exactly reproducible to about 2 percent, which turns a debugging luxury into something labs can afford to leave switched on.