The Sound of Silence: What Your AI Agent Does Not Tell You. Ha, seems like this guy just discovered the importance of having a large context window. Or the value of system prompts. Yeah, when you just have a conversation with an agent, it will soon forget what you talked about before. The oldest inputs fall off and are replaced with new inputs. If you want it to remember certain instructions, they must be put into a system prompt which gets loaded every time.
––– 良くない–––
Mistral debuts Large 4 ‘Le Chonk', a 1-trillion parameter text output model. Everyone has been talking only about American and Chinese models. How we've forgotten that France has a model, too. Now, Mistral announces "Le Chonk" with a trillion parameters. Yeah, we're not going to be downloaded that beast on our machines. It seems to be at pace with the best of the Chinese models, but the Chinese models are smaller. Wake me up when Le Chonk loses weight or gets much better.
––– 凄い –––
Effects of creatine supplementation with and without exercise and diet intervention on body composition, cognitive function, and markers of health in middle-aged and older adults. So you don't have to exercise to get the muscle-retaining benefits from creatinine, being a senior. I used to find it hard to take creatine. The large capsules or the inconvenient powders were unappealing. But I've discovered the chewable tablets by Bodytech. They're quite tasty and easy to use.
––– 凄い –––
iPhone Duo display confirmed to peel off on purpose. Wow, the matte nano-texture screen covering on the iPhone Duo is replaceable. It's there to preserve the crease-free hinge. Very nice. Apple thinks of these things, and now others will probably copy this.
––– 凄い –––
For Robertson, currently studying for a Ph.D. in quantum-safe cryptography at the University of Amsterdam, the problem is even more acute. She has seen all of her mathematics colleagues turn to using LLMs intensively in their research, with many maxing out their Pro subscriptions, and some even spending thousands of euros on additional tokens: “And these are Ph.D. students who don’t have thousands of euros,” she added.
Robertson said that there are colleagues who feel coerced by the tech giants, with the likes of OpenAI announcing in July that it is giving away 100,000 free licenses to its frontier models for researchers in academia. And there are other colleagues who feel they simply have no choice: “If you don’t work at the rate at which you could work with LLMs then you will be behind your peers who will be applying for the same jobs as you.”
Mathematics has changed irreversibly. Now, mathematicians have to get access to the best frontier models, and learn how to design prompts.
––– 凄い –––
Newborn Genome Sequencing To Be Offered In Oregon. It's just going to be available at Peace Health and OHSU, but Providence and Legacy will probably add it on, too. I don't think it's something every baby should have done. Only for those where you suspect an abnormality, and want to get prepared. The asymmetry in this information is clear: there is no mutation that is good news. The only good news is no harmful mutation, but many mutations are bound to show up. When we don't know what the implications of the mutations are, we just call them polymorphisms – just a variant of some sort. But it's not worth the anxiety and heartbreak, unless there is a mutation for which you can take action. Which is very rare. And the action is usually to remove some organ as a precaution.
––– 凄い –––
If I were Adobe, I'd be a little concerned:
––– 凄い –––
