Hacker News

in my book anything that is (I'm sure well intentioned) and injected to help the agent — but doesn't help it — is a waste of tokens

but even injected context that when I read it sounds useful can oversteer the model and make it second guess or take a more complicated route than it normally would

(you can see this when looking at traces with and without that injected context)

often harnesses also mention in their system prompt locations of markdown files that the model can consult if the model thinks they might help

that hint alone as part of the system prompt can be strong enough to make the model read in more tokens than would have been necessary

'spam' is maybe a harsh way to say it

unfortunately I don't see an easy way other than to invest time and tokens into finding out which parts of the added context (in system prompt, injected in turns etc etc) are actually helpful or harmful and when

I'm just doing the easiest thing I could think of: start from nothing or close to nothing

that seems to work better than what most harnesses are doing

turns out GPT 5.6 Sol is all you need

> Claude Code injects a ton of tools into the system prompt

It's so unfortunate they don't let you use the subscription with other harnesses anymore - since even if I used OpenCode they'd still get a bunch of useful data from the API calls, meanwhile I could stretch their tier limits way further.

The system prompt part is surely cached across all users.

ty re --disallowed-tools

for coding agents 'shell' is often all you need (just make sure the environment has the necessary tools)

appreciate the tip, kind netizen

smol only has 1 tool: sh

the system prompt of smol is shorter than the system prompt of Pi

smol has no system prompt

system prompt of Pi 0.83.0

""" You are an expert coding assistant operating inside pi, a coding agent harness. You help users by reading files, executing commands, editing code, and writing new files.

Available tools: - read: Read file contents - bash: Execute bash commands (ls, grep, find, etc.) - edit: Make precise file edits with exact text replacement, including multiple disjoint edits in one call - write: Create or overwrite files

In addition to the tools above, you may have access to other custom tools depending on the project.

Guidelines: - Use bash for file operations like ls, rg, find - Use read to examine files instead of cat or sed. - Inspect PI_* environment variables for current model and session details. - Use edit for precise changes (edits[].oldText must match exactly) - When changing multiple separate locations in one file, use one edit call with multiple entries in edits[] instead of multiple edit calls - Each edits[].oldText is matched against the original file, not after earlier edits are applied. Do not emit overlapping or nested edits. Merge nearby changes into one edit. - Keep edits[].oldText as small as possible while still being unique in the file. Do not pad with large unchanged regions. - Use write only for new files or complete rewrites. - Be concise in your responses - Show file paths clearly when working with files

Pi documentation (read only when the user asks about pi itself, its SDK, extensions, themes, skills, or TUI): - Main documentation: /usr/local/lib/node_modules/@earendil-works/pi-coding-agent/README.md - Additional docs: /usr/local/lib/node_modules/@earendil-works/pi-coding-agent/docs - Examples: /usr/local/lib/node_modules/@earendil-works/pi-coding-agent/examples (extensions, custom tools, SDK) - When reading pi docs or examples, resolve docs/... under Additional docs and examples/... under Examples, not the current working directory - When asked about: extensions (docs/extensions.md, examples/extensions/), themes (docs/themes.md), skills (docs/skills.md), prompt templates (docs/prompt-templates.md), TUI components (docs/tui.md), keybindings (docs/keybindings.md), SDK integrations (docs/sdk.md), custom providers (docs/custom-provider.md), adding models (docs/models.md), pi packages (docs/packages.md), environment variables (docs/environment-variables.md) - When working on pi topics, read the docs and examples, and follow .md cross-references before implementing - Always read pi .md files completely and follow links to related docs (e.g., tui.md for TUI API details) Current working directory: /workspace """

I will look into it more to see if I have configured it wrong but I think the token efficiency hurts cache use as caching only starts at 1024 tokens so for tasks where smol is under or close to 1024 tokens most of them are uncached

I built an ad-hoc custom comparison framework to inspect system prompts, caching behaviour, tool call outputs, exact api requests and responses and so on

I agree fewer tokens is not necessarily better but a bit counter-intuitively often the harness using fewer tokens is not only done faster but has better results

(that said: of course check the results, look at the full traces, agree!)

smol is basically this 9 line python agent re-implemented in Go

https://news.ycombinator.com/item?id=49006862

I'll have more about it in the next hours/days, you can follow me on twitter in the meantime (https://x.com/__tosh)

smol is also prefix caching

the uncached tokens are also from runs where smol finished a task below 1024 tokens (the minimum amount of tokens needed to activate caching) which is less tokens than other harnesses are using for their system prompt (!)

> GPT-5.6 and later models: Caching is available for prefixes containing at least 1,024 tokens. This is a strict minimum.

https://developers.openai.com/api/docs/guides/prompt-caching

so in this specific case the count of uncached tokens for smol makes it look worse than it actually is

that said: it does makes sense to add more tasks that are difficult enough to fill the context window to compare the harnesses for how well they deal with compaction

staying below compaction (or with compaction at fewer compactions) is not only cheaper and faster, it also helps the agent stay on track

the tasks were all simple agentic tasks

like creating a checksum of a file, merging csvs and so on, fixing a makefile pipeline

with known 'good' outcomes

all harnesses could reach the outcomes, only cost, time, number of tool uses and so on were different

(Claude Code failed once in 1 task but I think that was just an unfortunate outlier, the tasks aren't that difficult)

agree, that makes it a bit tricky to compare (esp if you also want to add different models and reasoning levels into the mix)

I will add more tasks (esp longer ones) and think more about grading, the current tasks were easy to grade because the desired outcomes are well specced but I will also look into more open ended tasks and how to grade those

thank you!

smol currently is very simple so it definitely does less things, like no subagent orchestration

I will look into how token usage looks like for longer sessions and more complex tasks

re caching: the cache ratio for this bench looks 'bad' for smol because it often finishes a task before caching kicks in (caching starts at 1024 tokens)

thank you for flagging this

> faster than Claude will take money out of your pocket and into theirs

Damn, and I thought Anthropic was fast.

You can use VSCode / Zed, run any CLI coding agent in the integrated terminal and review changes using the Git working tree diff view the IDE offers. Extensions seemingly give more ergonomy, like diffs of agent changes separate from Git, or option to comment on specific file or section, but I found out that agent CLIs don't really need that today -- most of the time they know what I mean without needing to explicitly point at code locations.

Can't VS Code extensions solve your use case?

You can use Intellij Ides, at least free ones.

I also want to believe that Cursor team working on harnesses for much longer then openAI and Anthropic can deliver better results overall. Claude code is being pretty much vibe-coded slop doesn’t give a feel of a good product.

Cursor has two “modes” agent and ide.

In agent mode it’s horrible for editing files etc but better if you are juggling multiple chats

I felt the same, but I realized, that if you vibe code more, the changes make sense. Simple example:

Should the AI sidebar be on the left, or on the right? At some point they swapped them automatically, which was jarring (maybe a bug). But now I realize that if I'm not the one primarily writing code and navigating it, then I prefer agents in the "primary" position on the left (esp. if I have the browser on the left half of the screen, the agents are in the middle).

What annoys me the most about Cursor is the random stuff that breaks (due to its nature of VS Code fork).

The tab experience is still unparalleled. The cost might be worse than Codex/Claude sub, but they are all much, much, much cheaper than API pricing, so depends on your POV.

Huh, never thought about that 2nd point - I'm also transitioning to pi, but didn't think to change to neovim as well. I guess your logic holds for claude code users as well though.

Sounds like on GitHub. That's how I do it too.

git diff

Does that mean you’re paying api prices + cursor markup (?), or can you somehow piggyback on your Claude subscription? Fable via the API is really too expensive

You can use Cursor CLI, it's a terminal agent, just like Claude Code and Codex. I use it sometimes, I haven't opened the actual Cursor IDE in ages.

The potential benefit of Cursor CLI (vs CC and Codex) is that you can easily between all major models (by Anthropic, OpenAI, xAI, as well as Kimi K3 and GLM 5.2). I found it useful when reviewing work - e.g. I implement using Opus then review using Sol, etc. Models by different providers tend to have different perspective on things and they can find different issues with the code.

It sounds like your wife might get left behind by you.

But doesn’t vscode’s settings.json have most if not all settings stored inside it? Should be a simple copy and paste.

I actually use grok as my planner and then composer for coding like opusplan in Claude.

It seems to me that people have a dislike like for Elon then just blanket everything with the same label.

It might be called grok out it was trained by the composer team using a large chunk of their training data.

Given that cursor has been bought by Musk, the nazi branding might be on purpose

please, you're dramatically overestimating who gives a shit about this.