I think I figured it out
Thursday, Nov 12 · 42 min · about 9 min to read
Theo-brain watched all 42 min
Theo argues that only two coding models matter right now: Opus 5.5 and Fable 5.2. Use Opus as the default; it is meticulous but bloats work and rarely pushes back. Reach for Fable when the work can't be objectively checked, when Opus's ideas feel samey, or when a stuck branch just needs someone to say it's fine to ship.
Skim itThe notes carry the decision rules; watch the two stretches below if you want his real-world stories behind them.
Tap any time to jump the video there.
Key takeaways
Theo says that, barring very specific reasons, using any model other than Opus 5.5 or Fable 5.2 for coding is a mistake. He admits he won't spend much time justifying this.
It is insane that we are at this point, but without very good and very specific reasons, using anything other than these two is insane. I could spend time
Split by whether the work can be verified. If you can check correctness and iterate toward it, like rewriting TypeScript in Rust, Opus is more than good enough. If 'better' is a judgment call needing nuance, creativity or exploration, Fable finds things Opus never would.
If it is possible to verify the work and you can hill climb your way there, whether this is something like rewriting TypeScript in Rust, you can verify that it's correct relatively easily. Opus is more than good enough for that. But if
it is a thing that you cannot objectively say this is better than that, if it is a thing that requires a little more nuance or creativity or exploration or curiosity, Fable is going to find things Opus would never. Any
Opus is the 'teacher's pet'. It almost never misses something, but it wants to please. Unless told otherwise, it quickly bloats a spec and lets feature creep take over.
homework. It is incredibly rare that Opus misses something, but it does have a negative effect from that. It goes too far sometimes. Since this model is so desperate to appease, it will, unless told otherwise, quickly bloat out a spec, allow for feature creep to take over. It's a little bit too quick to let
On the $200 Claude plan, Fable is capped at half your weekly limit. Theo says that works out to about $3,000 a month of Fable versus $14,000 of pure Opus, so less than a quarter as much Fable.
feel obligated to show it again. The 50% limit on your Fable gives you around 3 grand or so a month on the $200 plan. You get 14 grand of pure Opus. So it's less than a fourth as much Fable in your $200 plan. To be clear, you still have
Speed is not a reason to pick one over the other. Theo consistently sees Fable 5.2 at about 106 tokens per second, which is within the margin of error of Opus 5.5.
Fable 52 I consistently see 106 TPS. The speeds I am seeing in my day-to-day are within margin for error of Opus 55. Yeah. There is no speed difference from Fable to Opus that's real. There just
Two skills here are new to you.
Choosing an effort level for Opus
- What it is
- Effort level controls how long the model thinks and how much it checks before answering. Opus offers low, medium, high, extra-high and max.
- Why Theo cares
- He calls high 'the goat'. He rarely finds tasks it can't finish, it doesn't overdo things, and it isn't much less token-efficient than medium.
- How to try it
- Have your agent run Opus 5.5 on high as the default.
- If you stay on medium and it misses things, bump to high or extra-high.
- Avoid low and max; Theo calls both garbage.
medium and X high are both incredible. High is the goat. I have rarely found tasks high couldn't complete and it also doesn't do too much and it is not that much less token efficient. I have been very happy with high. Play with them
The 'suggest something else' prompt line New
- What it is
- A line added to prompts that gives Opus permission to push back: 'If you think this is a good idea, do it. Otherwise, suggest something else.'
- Why Theo cares
- Opus rarely challenges the task it's given. Theo says this one change meaningfully improved his outputs. He adds that it still sometimes suggests bad alternatives.
- How to try it
- Append that exact sentence to task prompts for Opus.
- Judge each counter-suggestion yourself rather than accepting it by default.
much better results from Opus when I started adding to my prompts. If you think this is a good idea, do it. Otherwise, suggest something else. That one small change with how I use Opus has resulted in meaningfully better outputs,
Opus builds the test setup, Fable audits it
- What it is
- Theo had Opus build a thorough test environment for a slow-loading problem in his app T3 Code. He then asked Fable to verify that work, probe for the real problems and propose fixes.
- Why Theo cares
- Opus built a great test setup but missed the root cause. Fable spotted it immediately, and the fix cut data sent over the network by about 150x in some cases.
- How to try it
- Have Opus build the reproduction and test environment for a hard bug.
- Then ask Fable to verify Opus's work, find the edges that actually cause problems, and propose solutions.
I moved over to Fable and said, verify the work Opus did here to create this environment to figure out and test all of these bad cases. Then I want you to play with it and find the edges that actually cause problems and propose solutions. And it immediately latched onto this exact problem was like,
one change was able to cut the amount of data being sent over the wire by like 150x for some of my edge cases. Fable
Haiku for bulk pattern-matching New
- What it is
- Haiku is Anthropic's smallest, fastest model. Theo uses it to sweep large amounts of data or every file in a codebase for a pattern that code alone can't detect.
- Why Theo cares
- He rejects Anthropic's advice to use Haiku for side tasks. He thinks its real strength is cheap, high-volume scanning.
- How to try it
- When you spot a bad pattern you can't search for mechanically, have your agent run Haiku over every file to flag it.
never ever ever use Haiku for a side task. I would use it for bulk analysis to go through a ton of data and have Haiku pattern match it or having a pattern you notice in your codebase that sucks that you can't programmatically identify. So having Haiku go through every single file and identify that,
Advice
If money is tight, buy the most expensive Claude subscription you can afford. Theo claims $100 on a Claude plan gets more work done than $300 on any other model, roughly $3,000 to $5,000 of usage.
subscriptions is insane. You will get more tokens and more work done for $100 in a Claude sub than you would get for $300 on literally any other model in the world because that $100 sub's going to get you like three to five grand of
If Opus has been going in circles on a thread you're trying to land, switch it to Fable. Fable may be smart enough to fix the problem, or it will tell you the work is already fine.
you should switch to Fable for two reasons. One, Opus might be stuck on something that it's not smart enough to fix and Fable is smarter, it might fix it. But more importantly, if it's already good enough, Fable will say you, it's fine. Wow, this video is a lot
If Opus on medium handles everything you give it, you aren't asking for enough. Give it bolder tasks.
that he's not yet found something that medium can't do. You got to start asking it to do bolder things then. That's the other take is if medium is enough for you, go bolder. You're not asking for enough yet. Addy's really trying to
When you want a list of possible solutions, ask Fable rather than Opus. Opus's options tend to repeat each other; Fable gives more variety, some better and some worse.
and two of them were very similar, one of them was a little different and okay and the other two just made no sense and it pissed you off, I'm I'm from a lot of experience here, Fable will give you a better variety of options, some of which might make a lot more sense, some of
Hot takes
There's no good reason to pick Sonnet 5.5 in the model picker right now.
His reasoning: In his test it cost more than Opus and produced a worse result. Without a clear path it burns tokens, because it's not that smart.
I don't know when you'd use Sonnet right now. It ended up being more expensive than Opus for fish slop and made a worse version. Sonnet is actually turned out to be one of the most expensive models when it doesn't have a clear path, which happens a lot cuz it's not that smart.
Max effort is garbage and you shouldn't use it.
His reasoning: Max doesn't just raise how much the model may think. It forces heavy thinking even on simple tasks, which leads to the model doing a lot of dumb things.
about this before. Max doesn't just change how much the model's allowed to think. It changes how little the model's allowed to think. It forces the model to think a lot even if the task is simple, which results in it doing a lot of stupid So personally I think
He wants his models to be lazy and cheat occasionally, which is what Fable does.
His reasoning: Lazy engineers who cut corners find genius solutions to get out of work.
overthinking it. Fable is kind of lazy. It's kind of willing to cheat. And honestly, I want my models to be lazy and cheat occasionally. I like engineers who are lazy and cheat occasionally because they will find genius solutions to get out of So how do we apply
How Theo thought about this
He picks between models the way he once picked human code reviewers. Allen knows the codebase deeply, so call him when a change is high-risk and must be perfect. Brad is the more experienced engineer who is new to the codebase; call him when you want to ship. Opus is Allen, and Fable is Brad.
would hit up Allen when I need it to be perfect, when the changes are really high risk, and I want to make sure nothing can go wrong in them. Allen is the person I call because he knows all
small. And I hit up Brad when I want to ship. Even though Brad is more experienced, even though Brad could hypothetically find more holes in the PR than Allen can, he has a different level
His own evals ranked Fable below Opus and Sonnet at code review for letting things slip, and at first he called that a flaw. Then he audited what Fable missed and found the gap didn't matter. So far, everything Fable called good to merge was.
better. I did a deeper audit of the comments that Fable leaves on things compared to Opus and Sonnet. The ones Fable missed, the reports Fable didn't file, as well as the less important things or the incorrect things Fable said, the gap doesn't matter. One of the
obviously like knock on wood, this could change, so far, everything Fable said was good to merge ended up being good to merge. I had a bunch of random that
Sign in to mark what you've watched and keep notes.
Have a question about this one? Ask about this video →
Theo Daily is unofficial and not affiliated with Theo or t3.gg. Written by Claude Opus 5.5 from each video's transcript.
