The Cognitive Revolution · 3 min instead of 2 h 1 m · Oct 10, 2026
Reasoning models enumerate options, then pick from the soup
Eric Bigelow
Researcher, Goodfire
Nathan LabenzHost · Host, The Cognitive Revolution
Founded AI video company Waymark and red-teamed GPT-4 before its release.
About this segment
Bigelow argues modern reasoning models fix the backtracking problem by listing every option, then choosing. He says reasoning is almost a misnomer for this.
From “How Agents Decide: Goodfire's Eric Bigelow on Critical Tokens, Phase Shifts, & In-Context Learning”, Oct 10, 2026. Watch the full episode
Takeaways
- Repeated 'wait' in DeepSeek R1 solved backtracking, not human-like reasoning
- Output diversity may come from the reasoning chain itself
- Raw capabilities were likely in the base models; RL sharpens the distribution
In this episode
Full episode- 6:30Resampling every token shows where answers collapse2:14
32:33Reasoning models enumerate options, then pick from the soup2:55
Transcript
Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.
- 32:38
learning leads to such catastrophic forgetting that the internal representation structure of the model radically changes, but I kind of, I kind of doubt that'll happen because that I think that will probably mean that the models will break. Like it may be that they, there is massive internal representational reorganization, but things are still structured and the model can still do lots of like all the old things it can do, but in new ways. But it seems to me more likely that if, if we're in a world where really there is this collapse of common internal structure, then the models are probably, probably breaking in that case, unless in In this world of continual learning, we're using an ungodly amount of compute for every one of these 10 million people to train them on, you know, do a lot of meta learning so
- 33:22
that they maintain their performance on all these tasks and learn entirely new representational structures. So yeah, I think in that world, it's kind of like having like a bunch of slight tweaks on a main model. And we probably will have like tools that we can directly translate to these fine-tuned models and maybe just do like a little bit of fine tuning on on on the tools that we have and the probes and such to the question about thinking about continual learning and in context learning my personal view i mean again i i don't have strong perspectives around their around the architecture argument but i think there's sort of a a bit of a a bitter lesson around like prompting and in context learning and context engineering and harnesses and all that that they might be all you need for quite a lot of stuff and i have a bias towards this because this is what i work on
- 34:14
but i kind of for me the most likely outcome in the in any short term is that we don't have custom models that are really trained on very different data, but they're actually, what's happening is it's different context, right? As you were saying, this is sort of what already happens. We have different LLMs that have different memories. And the memory is just the way I see it, a bunch of data that is tokens that are in context learned every time the model generates an answer, it sees this long memory.
- 34:42
And I think it's just, it's incredibly easy to get a lot of juice out of that without squeezing too hard. And then you end up with these memories that are pretty interpretable and uh yeah i think i think some variables would have to change in the equation for me to see it being really that promising to fine-tune models separately on individual people's data to have customization really pay off in that way and so maybe my my my personal best guess is that we're gonna see more context engineering and more of this this kind of thing more harnesses more memories and i think that's really what we've seen in the last year or so with with the rise of agents is is a lot of a lot of that engineering is just going