/Tech5h ago

T3 Stack creator Theo shares Fable AI workflow that restricts effort settings to avoid rate limits

Story Overview

T3 Stack creator Theo Browne walks through concrete ways to stretch Claude Fable 5 sessions in Claude Code without tripping rate limits, including locking effort settings to high, delegating implementation work to Codex, and feeding the model a running CLAUDE.md file that keeps context tight across long agentic runs.

3295.8K193#1 BOOKMARKED4.5K437.6K

3295.8K1934.5K437.6K

#1329

Original post

Theo - t3.gg@theo#1329inTech

I've been getting a TON done with Fable today and I'm not hitting rate limits. Wanted to share some tips on how I'm doing that

1. I only use Fable on "high" effort for now. xhigh is token hungry. max/extra is a furnace with worse outputs than lower options imo

2. I taught Claude Code how to use Codex as a fallback for lots of implementation tasks. GPT-5.5 is incredibly steerable, and Fable can learn how to steer it

3. I wrote up a big section in my CLAUDE[.]md on how to prioritize different models for different work when orchestrating workflows and subagents

4. Things that are unnecessarily token hungry (computer use, codebase analysis, etc), I do with other models and report results back to Fable

5:45 PM · Jul 1, 2026 · 324.1K Views

Developer Impact

Model hand-offs preserve headroom on premium tiers

Theo trains Claude to spot when a task is better routed to Codex, turning the higher-cost Fable instance into an orchestrator rather than the sole workhorse.

Cost Pressure

Context files and effort caps turn expensive runs sustainable

A persistent CLAUDE.md plus strict high-effort discipline lets builders keep multi-day workflows alive before credits or weekly allowances run out.

Sentiment

Positive users praise tips for efficient Fable AI workflows and call the $200 plan a solid deal for avoiding rate limits, while negative users complain about still hitting limits, extra costs, and token waste.

Pos

78.5%

Neg

21.5%

63 comments with sentiment.

Cluster Engagement

Digg Deeper

No Digg Deeper questions have been answered for this story yet.

Posts from X

Most Activity

VIEWS72.2KREPLIES78

Theo - t3.gg@theo

I've landed a dozen PRs and have more cooking now. Still haven't managed to hit rate limits on the $200 plan

Theo - t3.gg@theo

I've been getting a TON done with Fable today and I'm not hitting rate limits. Wanted to share some tips on how I'm doing that

1. I only use Fable on "high" effort for now. xhigh is token hungry. max/extra is a furnace with worse outputs than lower options imo

2. I taught Claude Code how to use Codex as a fallback for lots of implementation tasks. GPT-5.5 is incredibly steerable, and Fable can learn how to steer it

3. I wrote up a big section in my CLAUDE[.]md on how to prioritize different models for different work when orchestrating workflows and subagents

4. Things that are unnecessarily token hungry (computer use, codebase analysis, etc), I do with other models and report results back to Fable

4h72.2K870120

BOOKMARKS1.4KLIKES932RETWEETS28

Theo - t3.gg@theo

This is the relevant section of my CLAUDE.md

I'll be real - I haven't read it much. Just vibed out what I was looking for with Fable, and had it confirm it can use Codex for the things I care about

I still find Codex to be WAY better at computer use, verification of UI/UX work, and generally more efficient at execution on well spec'd work

Minmaxing this has genuinely been really fun for me and I'm loving the outputs I've been getting.

I was throwing away ~50% of my end-to-end agent-driven PRs before building this workflow. I haven't had to close a single one today :)

Theo - t3.gg@theo

I've been getting a TON done with Fable today and I'm not hitting rate limits. Wanted to share some tips on how I'm doing that

1. I only use Fable on "high" effort for now. xhigh is token hungry. max/extra is a furnace with worse outputs than lower options imo

2. I taught Claude Code how to use Codex as a fallback for lots of implementation tasks. GPT-5.5 is incredibly steerable, and Fable can learn how to steer it

3. I wrote up a big section in my CLAUDE[.]md on how to prioritize different models for different work when orchestrating workflows and subagents

4. Things that are unnecessarily token hungry (computer use, codebase analysis, etc), I do with other models and report results back to Fable

5h37.1K9321.4K

Nick Dobos@NickADobos

@theo @DanielleFong What’s the idea behind the thin wrapper subagent? That’s an interesting one

Theo - t3.gg@theo

I've been getting a TON done with Fable today and I'm not hitting rate limits. Wanted to share some tips on how I'm doing that

1. I only use Fable on "high" effort for now. xhigh is token hungry. max/extra is a furnace with worse outputs than lower options imo

2. I taught Claude Code how to use Codex as a fallback for lots of implementation tasks. GPT-5.5 is incredibly steerable, and Fable can learn how to steer it

3. I wrote up a big section in my CLAUDE[.]md on how to prioritize different models for different work when orchestrating workflows and subagents

4. Things that are unnecessarily token hungry (computer use, codebase analysis, etc), I do with other models and report results back to Fable

4h7.5K132

Theo - t3.gg@theo

@NickADobos @DanielleFong Means that it works in workflows and Fable is not involved until work is done

Nick Dobos@NickADobos

@theo @DanielleFong What’s the idea behind the thin wrapper subagent? That’s an interesting one

3h6.9K172

Theo - t3.gg@theo

So far today...

T3 Code - overhauling onboarding [wip] - mobile app touch ups [wip]

Lakebed - object storage [merged] - centralized auth [merged] - whitelisting users [merged] - favicon support [merged] - proper observability [reviewing] - pricing model [planning]

Postplan (my html hosting for html plan files) - actual dashboard for managing plans [merged] - version control and repo linking for plans [merged] - proper auth end to end + cli auth with tokens [merged]

Fleet (management layer for all my dev boxes) - actually centralize all the config I have floating around machines [merged] - centralized skill and config management [wip]

I think this is most of it? Been a busy day I won't lie.

4h584132

simonrowland@simonrowland

@theo @DanielleFong Try my tool for orchestrating other agents from claude:

http://Github.com/simonrowland/goal-flight

I’ve been getting a lot built with it.

4h37213

Jonny@jonnyboy

## Picking the right models for workflows and subagents

Rankings, higher = better. Cost reflects what I actually pay (OpenAI has really generous limits), not list price. Intelligence is how hard a problem you can hand the model unsupervised. Taste covers UI/UX, code quality, API design, and copy.

| model | cost | intelligence | taste | |----------|------|--------------|-------| | gpt-5.5 | 9 | 8 | 5 | | sonnet-5 | 5 | 5 | 7 | | opus-4.8 | 4 | 7 | 8 | | fable-5 | 2 | 9 | 9 |

How to apply: - These are defaults, not limits. You have standing permission to override them: if a cheaper model's output doesn't meet the bar, rerun or redo the work with a smarter model without asking. Judge the output, not the price tag. Escalating costs less than shipping mediocre work. - Cost is a tie-breaker only; when axes conflict for anything that ships, intelligence > taste > cost. - Bulk/mechanical work (clear-spec implementation, data analysis, migrations): gpt-5.5 — it's effectively free. - Anything user-facing (UI, copy, API design) needs taste ≥ 7. - Reviews of plans/implementations: fable-5 or opus-4.8, optionally gpt-5.5 as an extra independent perspective. - Never use Haiku. - Mechanics: gpt-5.5 is only reachable through the Codex CLI — `codex exec` / `codex review` (my ~/.codex/config.toml defaults to gpt-5.5). Use the codex-implementation, codex-review, and codex-computer-use skills; for work they don't cover (investigation, data analysis), run `codex exec -s read-only` directly with a self-contained prompt. - Claude models (sonnet-5, opus-4.8, fable-5) run via the Agent/Workflow model parameter.

Using gpt-5.5 inside workflows and subagents (the model parameter only takes Claude models, so use a wrapper): - Spawn a thin Claude wrapper agent with `model: 'sonnet', effort: 'low'` whose prompt instructs it to write a self-contained codex prompt, run `codex exec` via Bash, and return

4h8422

Theo - t3.gg@theo

@aarush_kukreja No I wrote my own. This seems cool though

4h4582

Aarush@aarush_kukreja

@theo https://github.com/openai/codex-plugin-cc is this what you use

4h2731

Theo - t3.gg@theo

@RhysSullivan I have no idea tbh, it's probably using whatever I have set already (which is xhigh fast iirc)

I never come close to hitting my Codex limits so I don't care as much

4h72010

Noah Storey@NtrpyDev

@theo just hit my limit on my 2nd pro plan, got A LOT done and your tips about orchestration have helped a large amount on my token usage.

4h1271

Umesh Maurya@R_umeshmaurya

This is the real meta right now.! Most people are still trying to force one model to do everything and burning through limits + getting inconsistent results. Smart model routing + using the right model for the right subtask (especially Codex as fallback) is where the biggest productivity jump is happening... The CLAUDE.md prioritization section is probably the highest leverage part of what you built.

4h7821

cras clogu@over_loa

@theo I thought you didnot like fable? I havent tried it yet but gpt5.5 is still my go to for now.

4h1631

Deepak Sangle@_deepakss_

@theo How does fable orchestrate with gpt ? It's not some other orchestrator like openrouter is it?

4h10211

Rhys@RhysSullivan

@theo what thinking level are you doing 5.5 on?

4h7093

Theo - t3.gg@theo

@WarrenInTheBuff It you don’t set it up yourself you won’t understand it imo. Much better to just open up that .claude folder and start tinkering :)

2h1161

🇺🇾 Emilio@akade_dev

@theo expand on 3 pls? is it tied to the domain or more general dev related?

btw it's insane this has to still be manually

5h4832

MΞRCY DΞ GRΞAT 🦇🔊@MercyDeGreat

@theo What are you building?

5h7431

Sam Elliott@samelliottsam

@theo 2. I taught Claude Code how to use Codex as a fallback for lots of implementation tasks. GPT-5.5 is incredibly steerable, and Fable can learn how to steer it

How did you do this?

4h2781

Theo - t3.gg@theo

@RhysSullivan 5.5 is slower? Not my experience at all

4h2141