643 karma · joined June 1, 2021
In long editing sessions with multiple iterations, the context can accumulate stale information, and that actively hurts model performance. Compaction is one way to deal with that. It strips out material that should be re-read from disk instead of being carried forward.
A concrete example is iterative file editing with Codex. I rewrite parts of a file so they actually work and match the project’s style. Then Codex changes the code back to the version still sitting in its context. It does not stop to consider that, if an external edit was made, that edit is probably important.
EDIT: It's specifically GPT-5.4 High in the Codex harness.
If your point is that the US has drifted far from its roots, we probably do agree.
The 2x allowance last month was not of the goodness of their hearts. It’s pumping up the demand AND the revenue simultaneously. The squeeze before an IPO.
How many of us have looked into moving to the next tier during the last 48 hours? I know I have both professionally and for my private use. Moving 20% of $20 customers to $100 increases the revenue for the group with 1.8x.
- No, it isn’t.
Is it intelligence to attribute this quote to Monty Python, or isn’t it?
Remembering things is traditionally considered a part of intelligence, memory that is. The subject has been discussed for ages. Personally I favor the statistical definition based on the observation that given a range of different intelligence tasks, individuals that succeed in one task generally succeed in other tasks. There is shared variance, a g-factor. Intelligence.
Brought to you by Carl’s Jr.
Reading this as an avid Codex CLI user, some things make sense and reflect lessons learned along the way. However, the patterns also get stale fast as agents improve and may be counterproductive. One such pattern is context anxiety, which probably reflects a particular model more than a general problem, and is likely an issue that will go away over time.
There are certainly patterns that need to be learned, and relearned over time. Learning the patterns is sort of an anti-pattern, since it is the model that should be trained to alleviate its shortcomings rather than the human. Then again, a successful mindset over the last three years has been to treat models as another form of intelligence, not as human intelligence, by getting to know them and being mindful of their strengths and weaknesses. This is quite a demanding task in terms of communication, reflection, and perspective-taking, and it is understandable that this knowledge is being documented.
But models change over time. The strengths and weaknesses of yesterday’s models are not the same as today’s, and reasoning models have actually removed some capabilities. A simple example is giving a reasoning model with tools the task of inspecting logs. It will most likely grep and parse out smaller sections, and may also refuse an instruction to load the file into context to inspect it. The model then relies on its reasoning (system 2) rather than its intuitive (system 1) thinking.
This means that many of these patterns are temporary, and optimizing for them risks locking human behavior to quirks that may disappear or even reverse as models evolve. YMMV.