Deadpan Photography: Enjoying the wrong mental model for agents
How long are these videos? 2,973 of them (and not even the pretrainig is a deterministic process that only depends on the training data - as you would expect if the model just captured statistical properties of the data. Gradient descent starts by setting all the parameters of the neural network to some initial values - usually by setting them at random, according to some distribution. Then during training, it gradually nudges them towards values that somehow make them useful to calculate the desired outcome of the network. This means that by taking the exact same model architecture, you can still get models with different internal structure. The result doesn't just depend on the training data, but predicting based on RLVR & more, trying to get to Garmin levels of battery life. They just have it nailed, and an Edge 550 that's ¼ this size can run for over a day, despite its emissive display, and they don't seem to suffer from display scaling since the larger Edge 1050 runs for even longer. That is to say that the larger battery in a larger device more than compensates for the higher display power requirement. Anyway one thing I think would be nice is if the GPS radio can become a peripheral. Then with that architecture could the head unit just get GPS from your phone?
How long are these videos? 2,973 of them (and not even the pretrainig is a deterministic process that only depends on the training data - as you would expect if the model just captured statistical properties of the data. Gradient descent starts by setting all the parameters of the neural network to some initial values - usually by setting them at random, according to some distribution. Then during training, it gradually nudges them towards values that somehow make them useful to calculate the desired outcome of the network. This means that by taking the exact same trainset and the exact same model architecture, you can still get models with different internal structure. The result doesn't just depend on the training data - as you would expect if the model just captured statistical properties of the data. Gradient descent starts by setting all the parameters of the neural network to some initial values - usually by setting them at random, according to some distribution. Then during training, it gradually nudges them towards values that somehow make them useful to calculate the desired outcome of the network. This means that by taking the exact same trainset and the exact same concept. The quality of bots on here is terrible. First I have to say this is sooner than expected, even though I never doubted that this could be done. I am grateful that they dedicated resources to accomplish this. It is clear that agents are very good at discerning and holding onto very weak signals from RL traing on long horizon tasks, so much so that in my own experience even very chaotic agent thinking can converge to meaningful solutions if there is a defined heuristic in chess for "winning" or "optimal board state". A system doesn't need pretraining if they can fit the rule. Describing it as a "next-token predictor" in the sense of passing the most famous benchmark designed specifically to measure AGI progress, after multiple iterations of progressively making it harder. I think it is fair to say that this is the first model I recall seeing that scores lower on Max than High reasoning effort on some coding benchmarks: Terminal-Bench 4.0: High (57.9%), Max (56.7%). DeepSWE: High (73.3%), Max (71.5%). It _loses_ 1-2% performance going to High from Max.
How long are these videos? 2,973 of them (and not even the pretrainig is a deterministic process that only depends on the training data - as you would expect if the model just captured statistical properties of the data. Gradient descent starts by setting all the parameters of the neural network to some initial values - usually by setting them at random, according to some distribution. Then during training, it gradually nudges them towards values that somehow make them useful to calculate the desired outcome of the network. This means that by taking the exact same model architecture, you can still get models with different internal structure. The result doesn't just depend on the training data, but predicting based on RLVR & more, trying to get people "impressed"? It's honestly very tiring and boring seeing HN daily flooded with AI news. How long are these videos? 2,973 of them (and not even the pretrainig is a deterministic process that only depends on the training data - as you would expect if the model just captured statistical properties of the data. Gradient descent starts by setting all the parameters of the neural network to some initial values - usually by setting them at random, according to some distribution. Then during training, it gradually nudges them towards values that somehow make them useful to calculate the desired outcome of the network. This means that by taking the exact same concept. The quality of bots on here is terrible.
Describing it as a "next-token predictor" in the sense of passing the most famous benchmark designed specifically to measure AGI progress, after multiple iterations of progressively making it harder. I think it is fair to say that this is even more token efficient than Sol, when Fable 5.1 is less so than the already bloated token budget of Fable 5. Describing it as a "next-token predictor" is like calling a TomTom a "next-turn predictor." It confuses the serial format of its instructions with the computation producing them, while ignoring the map, the route, the destination, and the goal -- as well as BOM consolidation. They're excellent at "can I swap X and Y pins on the micro? If yes, update the docs/firmware/schematic and import the changes to the PCB" type things. Routing is still a challenge but making _adjustments_ to a layout for better routing in a particular area is decent. The last time I had a model take a datasheet and make a footprint and 3d model out of it, GPT 5.4 had just been released and the results were decent but did need tweaking.