Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

So this is just delegating certain work to dumber models? I certainly wouldn't use Gemini 2.5 Flash (!!?) for code writing as suggested.

I've never had an issue with Codex or Claude reading massive files, they're really good at precise greps.



> I've never had an issue with Codex or Claude reading massive files

Reading files isn't a problem they want to solve. The idea seems to be using a cheaper model to "scout" for the intended code, instead of an expensive one that reads all the things (and spends more tokens / thinks about them).

I think this might be useful because Opus 5 especially tends to over-read. So this looks like an "LLM Bloom filter", telling "hey this is the code you might want to read".


I have a stage-gated workflow that prioritizes “premium” token efficiency (Fable.) and getting the most out of my subscription services. (Which boils down to Fable running carefully prompted deepseek-flash agent teams that defer back to the managing agent for any design decisions in most work.) As part of that workflow the manager uses cheap reconnaissance agents to burn their tokens in order to build relevant repo context, instead of the managing model’s. I’ve been doing this since they released Opus and it occurred to me that most of my pre-implementation phase token use was going right into the garbage bin with file reads that have to be done to find the relevant code, but are very wasteful.

There’s an added benefit that the manager’s focus on strategy and task decomposition before actually handling the user’s prompted task directly seems to be a very good way to interact with Claude’s Fable safeguards, and I haven’t had any refusals doing this.

And while I haven’t ran any numbers, I can get orders of magnitude more out of my claude subscription doing this, especially with deepseek-v4-flash being as good as it is for as cheap as it is.


It sounds like you're halfway to gas town already... Not saying that is definitively bad, but I have avoided this myself since I don't want to get bogged down in trying to figure out the optimal multiagent setup.


I absolutely am, and it's completely a "building a better hammer" thing.


I also currently run multiple Claude sessions with Fabel as the brain coordinating the manager sessions which in turn spawn sub agents.


> I have a stage-gated workflow

This is a Claudism, right? I feel like I never saw "gated" used this way before it.


A normal person would say “my workflow has stages” and their normal coworkers would say “no kidding”.


I'm not normal, nor do i have coworkers. Sorry :(

Just a guy trying to make his subscription last longer than the single Fable prompt anthropic includes for 100 bucks a month, lol.


This sounds exactly like what Repoprompt was built for: https://repoprompt.com/

The community edition was open sourced when the creator got hired by OpenAI a few months ago.


The creator (Eric Provencher) worked for Unity before that and he is exploring game development tooling etc (with an open token budget) its fun to see where things are heading in the on-demand future of handsfree blender output and animation


> "LLM Bloom filter"

very good way to put it.


It's not a great analogy, since Bloom filters are guaranteed to not have any false negatives, only false positives.

That property would be very useful here, but I don't see how it would be achievable using LLMs.


still a great analogy even if not techincally equivalent


"a filter" is an even better analogy because it is also technically correct


thats not a analogy, its just a description


Pretty sure claude code already delegates reading a large codebase to haiku subagents.


As of July, the explore agent inherits the parent model, capped at opus.

So fable and opus use opus to explore. Sonnet uses sonnet.

I replaced my built in explore agent with one hardcoded to sonnet low effort.

https://github.com/anthropics/claude-code/issues/72940


It did used to use Haiku but that model is now too too far behind…


Im now fascinated by how exactly Bcherny generated that message using claude. Like is the entire issue resolution automated, or still guided in someway?


GH copilot as well, explore subagent is configurable.


> So this is just delegating certain work to dumber models? I certainly wouldn't use Gemini 2.5 Flash (!!?) for code writing as suggested.

Why not, though? I started using OpenCode + GitHub Copilot, but I burned through my Claude Sonnet quota in just three days. I switched to GPT-5.4-mini, which uses far fewer tokens, and it’s often just as good as Sonnet. I think optimizing token usage is a good exercise. We often assume a model will be terrible, when it really isn’t.


My code being correct 99% of the and costing 5x more tokens is vastly preferable to my code only being correct 90% of the time and using fewer tokens.


This is true, but with the newer generation of models you don't want to do this yourself, they're really good at orchestrating and triage. Run Fable or Astra on low/medium, and tell them to come up with a plan then direct subagents using a weaker model (I like GPT 5.6 terra medium) to implement and verify, and review their work.


> and it’s often just as good as Sonnet. I think optimizing token usage is a good exercise. We often assume a model will be terrible, when it really isn’t.

“Often” doesn’t sound great. If the smaller model fails then I just wasted a lot of time and tokens.


Yes, this makes little sense. It looks like it's a way to avoid having Claude read or write your code.

And why stop at 90%? I have this one weird trick to reduce Claude Code token use by 100%: use a different harness and model!


I wish I could say this explains a lot about the state of Spotify's apps, but they have been that way since long before AI.


Am I missing something? Unless you have infinity money, would it not be very stupid to pay for "precise greps?"


It's a few hundred tokens. If you are worried about a few hundred tokens you are in trouble, forget about "infinite money."


Oh, so "programmers" really ARE losing the plot.

You kids don't get it, it's not about the tokens, it's about the principle of the thing. No self-respecting real programmer would accept the loss of even a few tokens over programmatic efficiency and cleverness.


This is basically exaclty what Cursor started doing when Composer was first released.

The app would start using it for exploration tasks, and then as it improved it became the default for writing code and tests too. You can change it of course, but I find it does a pretty decent job if you have a large model directing it.

The parent model of course checks the work, but most of the time the handoff is good enough that no edits are needed.

It's also pretty fast and cheap, firing off a bunch of sub-agents to explore different parts of the codebase is a regular occurrence for the way I work.


IME Composer 2.5 is too dumb for any serious coding. Grok 4.6 is twice as expensive (but still much cheaper than Claude Opus) and it does a much much better job.


Composer has improved in last few months. I relied heavily on it last month as ran out tokens and composer free credits were available for use.


This does seem to just be a subagents implementation.


Thank you, the article was so verbose I couldn't get the point just by skimming it. And it was this simple...


As I understand this is something similar to "anchors", tools to let agents avoid reading whole files.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: