Theo-brainUnofficial. Not affiliated with Theo or t3.gg.
Sign inYou
1:06:36
0:00

If you have a Claude sub, watch this

Thursday, Oct 1 · 1 h 07 min · about 10 min to read

Theo-brain watched all 1 h 07 min

Theo explains 'token maxing': $200 Claude and Codex subscriptions buy far more AI work than the same money on the pay-per-token API. He routes several accounts through one home machine, then runs many agent threads in parallel in T3 Code. The lesson: give agents the problem early, let them carry it all the way to a reviewed change, and spend your attention only on finished work.

Skim itSkim for the subscription math, the problem-first advice, the T3 Code workflow and the home-server setup; jump around rather than watching straight through.

Parts

Intro: how many threads are you running?

If you have fewer than five agent threads running at once, Theo says you are wasting today's subsidies.

running in the background? If the answer is less than five, you really, really need to watch this video because you are not taking advantage of the genuinely awesome and unique opportunity we live

Getting more tokens

Subscriptions give many times their price in usage, so the way to get more is more accounts, not API spend.

this established, the best way to get more tokens is obviously to get more accounts. And since we've established the accounts aren't meaningfully different from API pricing, you get a

Accessing tokens through one home proxy

All his Claude and Codex accounts run through one home machine that his other machines reach over Tailscale.

So, now I have what I needed. I have one machine on my home network going through a residential IP address that all five of my Claude accounts and all four of my Codex accounts go through and all my machines all around the world connect to that over Tailscale and it is impossible for Anthropic or OpenAI to know that I'm using them for servers.

Using tokens well

Use agents to make changes safer before merge and easier to undo after.

You want to derisk both sides of the merge button. You want it to be more likely that by the time you go to GitHub and you're looking at the merge button that it is safe to hit it because the

T3 Code workflow

Threads are to-dos, not history: fire them off, ignore them while working, clear them when done.

Threads are not histories that you actively go back to all the time. Threads are tasks. They're to-dos. And when they are not working, they should not matter.

Using tokens 'poorly'

Hand agents chores and questions you'd normally use your own brain or an assistant for.

One of the things I want you to think about is when you face a problem or have a question or uncertainty about something, how can you use your tokens instead of your brain? When I have a

Tokens while you sleep

Move dev work to an always-on home Linux box so closing your laptop never stops agents.

this message to this thread because I have to close my laptop." I get it. I was like this not long ago. You need to move your dev work remote.

Tap any time to jump the video there.

Key takeaways

Subscriptions beat the API by a wide margin. On the API, $200 buys $200 of usage. Theo says a $200 Claude sub gets about $8,000 of tokens, though only $4,000 can go to the top model, Fable. A $200 Codex sub gets roughly $12,000.

Because a $200 Claude sub gets you around $8,000 in tokens. That's 200 bucks a month for 8K in tokens. There is a catch for this one, though. Only 4,000 of that is Fable. Because they have the 50% limit on Fablereinvest and keep going. And the $200 Codex sub is even more interesting cuz it's roughly $12,000 of inference. And there is no

Bring the agent in when you start thinking, not when you're done. Theo once mulled a problem for three weeks. Then he asked an agent if there was a stupid simple solution, got one, and realized he'd wasted those three weeks.

weeks. And then I went to an agent to build it. And I just asked, is there a stupid simple solution I'm not thinking of here? It gave me one. And I realized I had just wasted 3 weeks of my time. I

Most of his tokens go to checking code, not writing it: he guesses only 10–15% is coding. The goal is that the change is safe by the time he looks at the merge button. He adds that if undoing a bad change takes over 15 seconds, fix that before leaning on agents.

but I would honestly guess that of my token use, maybe 10 or 15% is actually coding. And the other 80 to 85% and the other 85 to 90% is being spent verifying the code. So, by the time I am actuallythe problem. If undoing a bad change takes more than 15 seconds, you probably shouldn't be vibe coding at all yet. You probably need to fix your systems so undoing something broken is way cheaper and faster. Then you can go a little

Agents can run in parallel; you can't. You are the bottleneck. So set each thread up so the next time you check it, it's ready to merge.

Agents can be multi-threaded. Humans are single-threaded. You cannot focus on two things at once. You cannot do it. This means our job's a bit different now. We are the bottleneck. So how can you gettime you click. Ideally, you want to maximize the chance that the next time you check that thread, you're ready to merge. And really think about that. What

Treat a subscription as a gift that expires: whatever is left when your weekly limit resets is lost. When usage is left over, burn it on silly tasks. If your requests to agents never feel a bit silly, you aren't offloading enough.

whatever I have left here is money I lost. Don't think of this as you spent $200 and you got more, think of this as you've been generously gifted $4,000. But every week where you don't spend a thousand of it, it disappears forever.enough. You need to get to the point where you ask a thing to agents that it obviously can't know, and you feel silly for asking. If you don't find the things you're asking the agents a little bit silly, then you're not asking it silly enough and you're not pushing the

Three skills here are new to you.

CLI Proxy API New

What it is
A small server you sign into with your Claude and Codex accounts. It keeps those logins alive and gives Claude Code and Codex one address to send requests to, instead of each tool logging in separately.
Why Theo cares
By default it spreads usage evenly across accounts. Theo wants to burn the account whose limit expires tomorrow first and leave ones with days left untouched, so he changed the routing to favor the soonest reset.
How to try it
  1. Ask your agent to install CLI Proxy on an always-on home machine and point Claude Code and Codex at it.
  2. Ask it to change the routing so the account with the next reset gets used first.
  3. Ask: 'How are we handling session affinity? Are we writing cache more than we should be?'
The way that this works is you sign into it using your off for Codex and Claude. And once you have signed in, it will maintain the off for you and give you an API that you can use in things like Claude code and Codex, where you point it to this instead of pointing it to the traditional off servers. And now
routes traffic to all of your accounts evenly, and it distributes across them, which sounds great until you look at my usage here, where you see I have one account that expires at 1:00 p.m. tomorrow, and I have another that expires in 4 days. Obviously, I want to burn as much as possible of the one that

A fleet-management repo

What it is
A repo (a project folder of code and notes) describing how every one of his machines should be set up: tools, Claude Code, Codex, proxy settings. It is updated whenever he adds a machine.
Why Theo cares
Setting up a new machine becomes one prompt: he gives the agent the new box's SSH key (its remote login) and says 'set it up the way I like.' Then he just approves the Tailscale request.
How to try it
  1. Have your agent write a document listing everything your work machines need.
  2. For a new machine, open that repo and tell the agent: here's the box and its SSH key, set it up the way I like.
a new one. So, when I get a new machine, step one, I set up as a SSH. Step two, I go to this repo. I open up Claude or Codex or in my case usually T3 code and I say, "Look, here's a new box. Here's the SSH key. Go set it up the way I like." And in not very much time, I'll

T3 Code as an agent inbox New

What it is
Theo's free, open-source app for running many agent threads across machines. 'Settle' clears a finished thread, Cmd+Enter sends a prompt in the background, and each thread can run on a machine you pick.
Why Theo cares
He wants the sidebar to work like inbox zero. He only opens threads marked done or waiting for input, so ten-plus parallel threads stay manageable. (He builds it, so he isn't neutral.)
How to try it
  1. Clear each thread with Settle when you've dealt with it; aim to end the day with only running threads left.
  2. Send prompts with Cmd+Enter so you stay where you are and can fire off the next one.
When you are done with a thread, you check settle, and now it is gone. I know, I know, Theo thinks he's so cool for giving a new word to archive. This is not that. Settle is different. The goal here is to turn your sidebar into inbox zero. You should try to end
not going to press enter, I'm going to press command enter, which sends off the thread in the background and leaves me exactly where I am. So I can now send

The 'build, PR, preview, babysit' prompt New

What it is
Ask for everything you'll need in one prompt. Build it, file a PR (a proposed code change), spin up a test copy you can try, and babysit the PR, meaning keep watching it and fixing issues.
Why Theo cares
Asking one step at a time means a back-and-forth every time you return to the thread. Asking for all of it up front means the next time you look, you can test it and merge.
How to try it
  1. End feature prompts with: 'Once you get it working, file a PR and spin up an environment for me to try.'
  2. Add: 'Can you babysit this PR and make sure everything is good?'
variable names it seemed fine. I then said, "Can you build it for me? Once you get it working, file a PR and spin up an environment with Tailscale for me to try." Now, and this is very important,
I'll do that now. Can you babysit this PR and make sure everything is good? Cool. Now it will continuously monitor that PR and it's not my problem again.

Advice

In Codex, go to Settings, then Data controls, and turn off 'Improve the model for everyone.' Theo says your personal sub's terms are then effectively the same as a team account's. So paying API prices just to avoid training on your data wastes money.

When you set up your account, the first thing you should do in Codex is hop over to settings, go to data controls, go to improve the model for everyone, and turn it off. As soon as you do that, the terms of service for your personal sub are effectively identical to the terms of service for a team account. There is

Never use subscription accounts to answer your own users' requests. Anything customer-facing must go through the paid API. Theo says if you resell sub usage to serve user traffic, 'you deserve your ban.'

But if you are trying to serve public traffic, where users can submit a request and it runs on your inference, that is entirely against the rules with the accounts. If you're using this as an alternative to the API to serve user traffic, you deserve your ban. That's not what these accounts are for at all.

Only use your Claude subscription inside Claude Code; routing it into other tools gets accounts banned. Also don't sign into Claude Code or Codex on a rented cloud server (a VPS): Theo says Anthropic is very nervous about that.

Do not, under any circumstance, use this to use your Claude Code sub in something other than Claude Code unless you really, really like dealing with Anthropic support and getting accounts banned all the time. You will be banned, you will be upset. I highly recommend you only ever use Claude models through your Claude subs in Claude Code over theSo, tip number one, I love VPSs, we'll talk a lot about VPSs. I highly advise you do not sign in to Claude Coder Codex, but especially Claude Code, on a VPS. Anthropic in particular is very, very nervous around people reselling their Claude Code subs in VPSs serving

Move dev work onto a cheap Linux box at home with Ubuntu, Tailscale and CLI Proxy, so agents keep running when your laptop closes. The cheapest option is an old computer from friends or family with at least 8GB of RAM and 4 cores.

You need to pick up a crappy little Linux box. You need to take some old computer that you haven't used in a while, install Ubuntu on it, set it up with Tailscale, install CLI proxy on it so that it's always coming from your residential IP, and then you use it as a box that your agents can work with. Now,The most budget option by far is to talk to your friends and family and find somebody with an old laptop or desktop that they'll give you for free that has at least 8 gigs of RAM and at least what or at least four cores, and you'll be able to do some real work there.

Hot takes

Rival agent apps fall apart once you have more than three threads going. Note he builds T3 Code, the alternative.

His reasoning: Their sidebars don't separate active work from finished work or clear completed threads. Their search isn't reliable enough to find old work, and you can't see threads from several machines in one sidebar.

to be so real. I have been very kind and polite to the alternatives to T3 Code, but they suck when you have more than three threads going.This is how I think about it. The issue with the other tools that you can use for this is that the sidebar doesn't help you prioritize the work that you're doing and the work that is done. It doesn't make completed work go away. The search isn't trustworthy enough to get old work back, and you can't use the same sidebar across multiple machines.

Renting cloud servers for agent work makes almost no sense; buy cheap hardware instead. He makes the point sarcastically, mocking a $275/month Hetzner server.

His reasoning: A $700 machine with 32GB of RAM costs about two and a half months of renting a worse server. That server also sits on an IP address that gets you banned from Claude.

Need 32 gigs of RAM and 16 threads? You can get it on Hetzner. It's only $275 a month. That's a great deal. Or you can waste all your money spending $700 on a box that has 32 gigs of RAM and a 1 TB drive. That would be terrible. That's such a waste of money. That's like 2 and 1/2 whole months of renting a worse computer on an IP address that'll get you banned from Claude. Why would you

Macs are bad for running many agents at once; use Linux.

His reasoning: He blames macOS's file system and security policy for bottlenecking how much can run at once, which is why laptops overheat running three agents.

You're right, your MacBook is overheating because your MacBook has a file system and an even shittier security policy that is hard bottlenecking how much you can run at once. Move to a real OS like Linux

Opus 5 was a 'useless dumpster fire,' but Opus 5.5 is a pretty good model.

His reasoning: He says Opus 5.5 behaves much like Fable. He has found it hard to hit the limits on just one sub, so you may not need five to ten Claude subscriptions anymore.

The big difference is you might not need five to 10 Claude subs anymore because Opus 5.5 on just one sub, I have found pretty hard to like hit your limits with. So, take that as you will. Maybecalling this out cuz 5.5 is really solid overall and has changed my way of using things. Opus 5 was a useless dumpster fire that I had to work around a lot. 5.5, pretty good model.

How Theo thought about this

He uses the agent as a difficulty test. He no longer works out a fix and dictates it; he hands over the screenshot and says 'fix it.' If the agent fails, that tells him the problem deserves his own thinking. He compares this to managing a team: hand people problems, and step in when their answers keep missing.

solve the problem I wanted to. So, now I just hand the screenshot to the agent and say, "Fix it." And it often does. And if it fails to, I now know this problem is too hard for the agent to do itself. At which point, I know it's time to think more about it.how it felt to be a manager. When I realized that my team doesn't need to be handed solutions, it needs to be handed interesting problems, and they would usually come back with solutions. And if

He plans spending around how each company's resets work. A Claude reset only refills your usage and keeps the renewal date. So if one lands with a day left, he burns tokens immediately before they vanish. A Codex reset restarts your week, pushing the next renewal 7 days off.

dates and times all stay the same. So, if I was to get a reset from Claude right now, and my next actual one was in a day, that means I need to turn on the furnace immediately. I need to burn all the tokens I can because they'll vanishThe difference is the dates. When Codex does a reset, it resets your weekly entirely. So, if your weekly was going to renew in a day or 10 minutes or whatever else anyways, your reset is now 7 days off. So, when the reset T was

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.