"I'm using OpenAI here"
proceeds to show a piece of code importing anthropic
was pretty confusing to me
How to code Claude Code in 200 lines of code
241–249 of 249 posts
Re: How to code Claude Code in 200 lines of code
#242Earlier quoted context omitted.
I run evals and the Todo tool doesn't help most of the time. Usually models on high thinking would maintain Todo/state in their thinking tokens. What Todo helps is for cases like Anthropic models to run more parallel tool calls. If there is a Todo list call, then some of the actions after are more efficient. What you need to do is to match the distribution of how the models were RL-ed. So you are right to say that "d…
Curious what kinds of evals you focus on? We're finding investigating to be same-but-different to coding. Probably the most close to ours that has a bigger evals community is AI SRE tasks. Agreed wrt all these things being contextual. The LLM needs to decide whether to trigger tools like self-planning and todo lists, and as the talk gives examples of, which kind of strategies to use with them.
I am taking for SWE bench style problems where Todo doesn't help, except for more parallelism.
Re: How to code Claude Code in 200 lines of code
#243Earlier quoted context omitted.
Models just naturally arrive at a conclusion that they are done. TODO hints can help, but is not infallible: Claude will stop and happily report there's more work to be done and "you just say the word Mister and I'll continue" --- this is a RL problem where you have to balance the chance of an infinite loop (it keeps thinking there's a little bit more to do when there is not) versus the opposite where it stops short…
> this is a RL problem where you have to balance the chance of an infinite loop (it keeps thinking there's a little bit more to do when there is not) versus the opposite where it stops short of actual completion. Any idea on why the other end of the spectrum is this way -- thinking that it always has something to do? I can think of a pet theory on it stopping early -- that positive tool responses and such bias it tow…
My pet theory: LLM's are good at detecting and continuing patterns. Repeating the same thing is a rather simple pattern, and there's no obvious place to stop if an LLM falls into that pattern unintentionally. At least to an unsophisticated LLM, the most likely completion is to continue the pattern.
So infinite loops are more of a default, and the question is how to avoid them. Picking randomly (non-zero temperature) helps prevent repetition sometimes. Other higher-level patterns probably prevent this from happening most of the time in more sophisticated LLM's.
Re: How to code Claude Code in 200 lines of code
#244Earlier quoted context omitted.
Curious what kinds of evals you focus on? We're finding investigating to be same-but-different to coding. Probably the most close to ours that has a bigger evals community is AI SRE tasks. Agreed wrt all these things being contextual. The LLM needs to decide whether to trigger tools like self-planning and todo lists, and as the talk gives examples of, which kind of strategies to use with them.
I am taking for SWE bench style problems where Todo doesn't help, except for more parallelism.
Was guessing that, coding tasks are a valuable but myopic lense :)
I'm guessing a self-updating plan there is sufficient. I'm not actually convinced today's current plan todolist flow makes sense - in the linked PLAN.md, it gets unified, and that's how we do ai coding. I don't have evals on this, but from a year of vibes coding/engineering, that's what we experientially reached across frontier coding models & tools. Nowadays we're mixing in evals too, but that's a more complicated story.
Re: How to code Claude Code in 200 lines of code
#245Can't believe Claude Code was launched less than a year ago. It's certainly a new tool in my kit and allowed me to approach and work on new coding problems in a new way.
Re: How to code Claude Code in 200 lines of code
#246I’d been wondering how conversation history actually works in these agent loops — the LLM itself has no memory, so whatever “history” exists is just text you keep feeding back in.
At a high level it seems to usually be one (or a mix) of:
- full transcript appended every turn
- sliding window of the last N turns / tokens
- older turns summarized into a rolling memory
- structured state (goals, decisions, progress) rendered into the prompt
- external storage + retrieval (RAG-style) to pull in only relevant past info
Under the hood I’m sure it gets more complex, but the core idea is pretty simple once you strip away the mystique: memory = prompt assembly.
Re: How to code Claude Code in 200 lines of code
#247Something I would add is planning. A big "aha" for effective use of these tools is realizing they run on dynamic TODO lists. Ex: Plan mode is basically bootstrapping how that TODO list gets seeded and how todos ground themselves when they get reached, and user interactions are how you realign the todo lists. The todolist is subtle but was a big shift in coding tools, and many seem to be surprised when we discuss it -…
ironic how the peak of code is a TODO list app
Re: How to code Claude Code in 200 lines of code
#248Re: How to code Claude Code in 200 lines of code
#249Earlier quoted context omitted.
https://github.com/jacobsparts/agentlib See CodeAgent or subrepl.py if you're just interested in the REPL orchestration. I also have a Python REPL MCP server that works with CC. It isn't published, but I could share it by request. My favorite part of code agent is the /repl command. I can drop into the REPL mid session and load modules, poke around with APIs and data, or just point Claude in the right direction. Some…
I request you to publish it! :-D
I haven't had time to make a repo, so I'll just drop this here for you:
https://jacobstoner.com/subrepl_mcp.py
Just make sure you have the deps installed and add it to CC as a stdio mcp server. Tested on Linux only. I use it daily.