There are no easy ways to learn.
...n. Smaller mistakes that are less likely to knock you out of the game, a faster feedback loop to learn.
Topic: feedback loop
63 chunks · 41 episodes
Topic summary
?
A short read on the topic's time range, peak episode, and strongest associations. Use it as the quick orientation before drilling into examples.
- feedback loop appears in 63 chunks across 41 episodes, from 2023-10-02 to 2026-04-13.
- Its densest episode is Bits and Bobs 9/16/24 (2024-09-16), with 5 observations on this topic.
- Semantically it travels with llms, compounding value, and ground truth, while by chunk count it sits between infinite software and Anthropic; its yearly rank moved from #34 in 2023 to #60 in 2026.
Over time
?
Raw mentions over time. Use this to see absolute attention, not relative rank among all topics.
Observations
?
The primary evidence view for this topic. Sort it chronologically when you want concrete examples behind the larger pattern.
It's very important to not over-fixate on your toe hold.
... increasingly doesn't work outside of it. Getting to a toehold (bootstrapping a feedback loop) is everything, and yet make sure that any specific toehold you find doesn't become your everything.
Viability is much more precious than explainability.
...ey simply don't have enough space to incorporate all of the swirling interlayed feedback loops, many of which are outside of us, to know which ones will be strong enough to overcome other ones. But once you know it works, you can often figure ...
Learning happens in feedback loops.
Learning happens in feedback loops. But the loop has to be minimally viable to even be a loop in the first place. Versus an open path. Once it is a loop then you learn as you iterate....
One of LLMs' primary superpowers: they have human-level ability, but they never get bored.
..., because you can more easily tell if you broke something. This allows a faster feedback loop. Even imperfect tests are better than no tests. As long as the tests are generally green, so you don't develop a boy-who-cried-wolf insensitivity to ...
You're a king when you deal with Claude.
... loop. It makes you lean more into the thing you were already doing. You need a feedback loop to balance it out. English kings realized they needed oppositional agents. Agents competing amongst themselves to represent the king today, keeping t...
Dictator mode allows you to execute quickly, but breaks the feedback loop.
Dictator mode allows you to execute quickly, but breaks the feedback loop. All throttle, no steering wheel. You won't realize what you've lost until it's too late. This is a massive and fundamental tradeoff.
Karma is in some sense literally true, but it's more obviously true in smaller contexts.
...ct you, too. The quicker your actions affect you, the more obvious the indirect feedback loops. If it's not obvious, then it's an "externality", and you don't think about it. In a small community, the actions reflect back on you quickly. If yo...
"Move fast and break things" works great when there are no indirect effects.
... dangerous when there are lots of indirect effects (e.g. externalities, or long feedback loops that are hard to steer).
Longer feedback loops are inherently much harder to steer.
Longer feedback loops are inherently much harder to steer. The way we learn is by taking an action with a predicted effect, seeing how the outcome matched the prediction,...
The longer the feedback loop, the more annoying to use--even if the loop gives good results.
The longer the feedback loop, the more annoying to use--even if the loop gives good results. You get around the loop, excruciatingly, and go "no, no that's not it, here's new dir...
As you get more senior, the feedback loops go from one day to two years.
As you get more senior, the feedback loops go from one day to two years. The steering wheel is connected with tons of leverage, but tons of lag. Very very hard to learn in that environment. S...
A few random off-the-cuff reactions to OpenAI's strawberry model in no particular order:
...ing I think is most interesting is how hard it is to steer based on a very long feedback loop. Set it and then come back to it later... and if you realize "oh crap I forgot to include some relevant context" you have to do it again. Longer feed...
One reason the internet grew as quickly as it did was because the capex was at the edge.
... value is created. More indirect, less sure, a slower loop. Systems with longer feedback loops grow and adapt much more slowly.
In a close-ended system (Mediocristan) feedback loops take you towards an average point.
In a close-ended system (Mediocristan) feedback loops take you towards an average point. In open-ended systems (Extremistan) feedback loops take you away from an average point.
Hallucinated mini-apps are having a moment.
...ificant upside. Anthropic Artifacts are just interface sugar, but they make the feedback loop immediate and help give a gradient of learning. But hallucinated mini-apps today have two significant problems. 1) They can't safely work with your d...
The Ouija board effect: minor amplitude but consistently aligned phenomena lead to significant movement of the collective, as if by magic.
...f moving randomly, but the people moving it are interpreting it in a continuous feedback loop. As it starts looking like it's moving to a letter, if it makes sense, then it starts going there faster and faster as more of the people with hands ...
LLMs today make boring mistakes because they can't learn.
...ns will be used during the training of future iterations of the model, but that feedback loop is indirect and very long. Compare to for example some search ranking tweaks that can update nearly instantly. Humans are not like this. They quickly...
In the Github Copilot feedback loop, if you don't like the recommendation, write a bit more comments and it will generate a new thing.
In the Github Copilot feedback loop, if you don't like the recommendation, write a bit more comments and it will generate a new thing. The next thing to do to steer it is fast and obvio...
Over-optimization is pushing deeper into a niche.
...ure that protects and makes you more efficient, but also traps you. A too tight feedback loop leads to homogenous over-optimized things that are prone to catastrophic population collapse. A too-tight feedback loop leads to over-optimization.