Theo puts Meta's Claude Code clone, Muse Code powered by Muse Spark 1.2, through benchmarks, a codebase audit, a game rewrite, and a live integration test to see if the price is the whole story.
Posted
1 weeks ago
Duration
Format
Review
sarcastic
Views
143.7K
3.3K likes
57 · 43
Big Idea
The argument in one line.
Meta's Muse Code trades benchmark-leading intelligence for extreme speed and, on its data-sharing 'contributor tier,' extreme cheapness, which makes it genuinely useful for fast throwaway analysis work but not yet trustworthy for real production coding.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
You want to know whether Meta's new coding CLI is worth installing today, not just its marketing announcement.
You're curious how a 10-20x cheaper 'contributor tier' pricing model works and whether the tradeoff, Meta training on your data, is worth it.
You use multiple AI coding tools already and want a sense of where a fast, cheap model actually earns a place in your workflow, like audits or summaries.
You're weighing raw model capability against speed and cost for narrow, well-defined tasks like PR review or codebase audits.
SKIP IF…
You're looking for a deep technical teardown of Muse Spark 1.2's architecture or training process; this is a hands-on user review, not a technical paper.
You need a tool you can trust today for unsupervised production coding; the video's own conclusion is that this isn't it yet.
TL;DR
The full version, fast.
Meta released Muse Code, its own Claude Code clone, powered by the new Muse Spark 1.2 model, and its benchmark scores land mid-pack, roughly tied with GPT-5.5 and Grok 4.5, well behind Opus 5 and Fable 5. What makes it interesting is price and speed: on the 'contributor tier,' where Meta trains on your data, it costs 10-20x less than standard pricing and runs about 191 tokens per second, roughly five times faster than comparable frontier models. Hands-on, it excelled at fast, narrow analysis work, auditing 222 open pull requests for 10 cents, or producing a codebase audit report in under a minute. But it struggled badly with sustained, open-ended tasks: a live attempt to build a working tool integration failed outright, and a blind model review scored its own integration plan 4.8 out of 10 against a rival model's 7 out of 10. Cheap and fast enough to be fun to experiment with, not yet reliable enough to trust with real production coding.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
01 · Cold open: Meta fandom and the Muse Code reveal
Theo explains his genuine respect for Meta's open-source track record (React, React Native, Llama) before revealing that Zuckerberg just announced Muse Code in beta, a Claude Code clone powered by the new Muse Spark 1.2 model, followed by a sponsor read for Greptile's T-Rex.
03:40 – 07:08
02 · Benchmarks and Meta's engineering culture
Theo reads Zuckerberg's announcement thread and the published benchmark charts, noting Meta only compared against Opus 5 and Sol 5.6 Terra, while explaining Meta's unusual internal engineering culture (custom Mercurial tooling, rewriting PHP into Hack) as context for how seriously Meta invests in its own stack.
07:08 – 09:40
03 · The pricing reveal: the contributor tier
After struggling to find current pricing info via Google, Theo finds the real numbers: standard pricing is $1.25/mil in and $4.25/mil out, but the 'contributor tier' cuts that 10-20x in exchange for Meta training on submitted data.
09:40 – 13:30
04 · Intelligence Index, speed, and cost per task
Theo walks through Muse Spark 1.2's Artificial Analysis Intelligence Index score (54, tied with GPT-5.5 and Grok 4.5), its output-token efficiency, and its OpenRouter throughput numbers (191 TPS average, up to 316 TPS best case), calling the speed 'absolutely nuts.'
13:30 – 20:10
05 · Hands-on: codebase audit vs Fable and DeepSeek
Theo installs Muse Code in the t3.chat codebase, asks it to investigate a suspicious event-sourcing pattern, and has it synthesize an HTML findings report using his personal Claude Code skill. He runs the same prompt through Fable and DeepSeek v4 Flash for comparison; Muse finishes first by a wide margin.
20:10 – 23:44
06 · Fish game rewrite and design site comparisons
Theo has Muse Spark 1.2 rewrite his prototype fish game in both 2D and 3D from scratch, finishing both in under five minutes combined versus over an hour with Opus 5, though the results are visibly broken (backwards-swimming fish, no mouse look). He also reviews Muse's design output on a model-comparison site and notes a distinct visual 'flavor' to its work.
23:44 – 27:40
07 · Investigating Muse Code as a provider
Theo asks Muse Code, using sub-agents, to investigate how it could be integrated as a model provider inside t3.chat, while separately having Fable investigate the same question and review Muse's resulting plan.
27:40 – 34:22
08 · Cost tracking and rate-limit chaos
Theo tracks running spend on the Meta Model API usage dashboard (from 13 cents up through several dollars), comparing it to roughly $32 spent on a lighter equivalent session with Fable and Opus, while repeatedly hitting rate limits on parallel sub-agent runs.
34:22 – 40:30
09 · The live integration attempt fails
After Fable reviews and Muse revises its integration plan, Theo tries to run the resulting code live inside t3.chat. It fails outright, never registering as a provider in settings. A blind third-model comparison (Sol) later scores Muse's plan 4.8/10 against Fable's 7/10.
40:30 – 44:42
10 · The PR audit test and final verdict
Theo's final test has Muse Code audit all of his open GitHub pull requests, spawning seven sub-agents; it reviews 222 PRs with mergeability scores and working links in under five minutes for 10 cents. Theo closes with his honest verdict on who should actually use this model.
Atomic Insights
Lines worth screenshotting.
Meta's Muse Code 'contributor tier' costs 10-20x less than standard pricing in exchange for letting Meta train on your usage data.
Muse Spark 1.2 scores 54 on the Artificial Analysis Intelligence Index, tied with GPT-5.5 and Grok 4.5 but behind Opus 5, Fable 5, and Kimi K3.
On OpenRouter, Muse Spark 1.2 averages 191 tokens per second, roughly five times faster than a comparable frontier model like Sol 5.6.
Auditing 222 open pull requests across a real codebase cost 10 cents on the contributor tier, versus tens or hundreds of dollars with a model like Fable.
A blind review by another model scored Muse's own integration plan 4.8 out of 10 against a rival model's 7 out of 10 for the same task.
Muse Code hallucinated an entire multi-minute investigation by confusing 'Muse' with Google's unrelated 'Antigravity' product before being corrected.
Rewriting a game prototype from scratch in both 2D and 3D took Muse Spark 1.2 under five minutes total, versus over an hour with Opus 5 for the same two tasks.
Muse Code automatically read the user's existing Claude Code personal rules and skills from the .claude directory rather than the emerging cross-vendor AGENTS.md standard.
Meta trained its coding model partly on internal engineers' actual usage of Anthropic's models, which one commentator joked makes it 'best in class at applying for jobs at Anthropic.'
Rate limits on the contributor tier hit hard enough to stall parallel sub-agent workflows, even after switching to the higher-priced standard tier.
Meta's own benchmark charts omitted its strongest coding models, Fable and Sol, showing only Opus 5 and Sol 5.6 Terra as comparisons.
A live attempt to make Muse Code operate as a working model provider inside another coding tool failed outright, never registering itself in the settings.
Takeaway
A coding model this cheap and fast still can't be trusted to finish the job.
PRICE VS TRUST
Muse Code's price and speed are real, but its reliability for actual coding work isn't there yet, so it earns a place for fast, narrow analysis rather than production coding.
01Cold open: Meta fandom and the Muse Code reveal
Meta's actual track record in open source, React, React Native, Llama, is legitimately strong, even though its AI coding models have lagged the industry.
Muse Code is explicitly built and marketed as a Claude Code clone, right down to reading the .claude directory's personal rules and skills instead of the newer cross-vendor AGENTS.md standard.
02Benchmarks and Meta's engineering culture
Meta's own benchmark charts only compared Muse Spark 1.2 against Opus 5 and Sol 5.6 Terra, leaving out its strongest internal models, Fable and Sol, entirely.
Meta's internal engineering culture, a custom Mercurial-based stack instead of Git, rewriting PHP into Hack, shows a company willing to rebuild core tooling rather than live with off-the-shelf limits.
03The pricing reveal: the contributor tier
Muse Spark 1.2's headline price is $1.25 per million input tokens and $4.25 per million output tokens, but that's not what most users will actually pay.
The 'contributor tier' cuts pricing 10-20x in exchange for letting Meta use your submitted content for training, effectively a free tier funded by data collection.
04Intelligence Index, speed, and cost per task
Muse Spark 1.2 scores 54 on the Artificial Analysis Intelligence Index, tied with GPT-5.5 and Grok 4.5, behind Opus 5, Fable 5, and Kimi K3.
On OpenRouter, Muse Spark 1.2 averaged 191 tokens per second, roughly five times the throughput of a comparable frontier model, which changes how it feels to use even when it isn't as capable.
On the contributor tier, Muse Spark 1.2 runs about 40 cents per task on the Intelligence Index benchmark, among the cheapest of any model tested.
05Hands-on: codebase audit vs Fable and DeepSeek
Muse Code produced a full codebase architecture overview and a synthesized HTML findings report in well under a minute, while a comparable Fable investigation was still running after four minutes.
Getting multiple models to independently investigate the same question and then cross-review each other's findings is a fast way to spot which model's reasoning actually holds up.
06Fish game rewrite and design site comparisons
Muse Spark 1.2 rewrote a small game prototype from scratch in both 2D and 3D in under five minutes total, versus over an hour each with Opus 5 for the same tasks.
Fast output doesn't mean correct output: the 3D rewrite shipped with backwards-swimming fish and broken mouse-look controls that took several more prompts to partially fix.
07Investigating Muse Code as a provider
Muse Code supports git worktrees, letting you test destructive or exploratory changes in an isolated copy without touching your main working directory.
Having a second model review a first model's plan before execution caught real gaps, and the reviewed plan scored far better in a later blind comparison, 7 versus 4.8 out of 10.
08Cost tracking and rate-limit chaos
Total spend for a genuinely heavy session of investigation, planning, and code changes stayed under a dollar on the contributor tier, versus roughly $32 for a lighter comparable session with Fable and Opus.
Rate limits on the contributor tier hit hard during parallel sub-agent workflows, even though sub-agent orchestration was explicitly marketed as a strength of the tool.
09The live integration attempt fails
After a plan was written, reviewed, and revised, the actual integration attempt still failed outright, never registering itself in the target tool's settings.
A model can be excellent at breaking work into parallel sub-agents while still being unreliable at seeing a single exploratory task through to a working result.
10The PR audit test and final verdict
Auditing 222 open pull requests, sorted by mergeability with confidence scores and working links, cost 10 cents on the contributor tier and took under 5 minutes.
For narrow, well-defined analysis tasks, audits, summaries, categorization, a fast cheap model can beat a slower expensive one on cost-to-value, even if it can't be trusted for open-ended coding.
A closed-source CLI that can't be integrated as a provider into other tools limits a fast, cheap model's usefulness to whatever its own first-party harness supports.
Glossary
Terms worth knowing.
Contributor tier
Meta's discounted Muse Code pricing tier, roughly 10-20x cheaper than standard pricing, offered in exchange for letting Meta use submitted prompts and code as training data.
Muse Spark
Meta's family of large language models; Spark 1.2 is the newest coding-focused release, and it powers Muse Code.
Intelligence Index
A composite benchmark score published by Artificial Analysis that ranks AI models on general reasoning and agentic task capability.
Terminal-Bench
A benchmark that measures how well a coding agent completes real command-line software engineering tasks.
ACP
A protocol layer some coding tools use to plug in different AI model providers without writing a custom one-off integration for each.
Worktree
An isolated copy of a git repository's files that lets an agent make risky or exploratory changes without touching the main working directory.
Mercurial
A source-control system Meta uses internally instead of Git, chosen for better performance at Meta's scale of code.
“They're effectively giving out the contributor version for free because they need training data so goddamn badly.”
sharp one-line explanation of the pricing strategy→ TikTok hook↗ Tweet quote
09:35
“Muse Spark is trained on the screen recordings of Meta employees, which makes it best in class at applying for jobs at Anthropic.”
quoted tweet punchline, instantly funny with no setup needed→ IG reel cold open↗ Tweet quote
39:00
“It's almost like it knows how to act like a modern smart model, but it doesn't know what the modern smart models know.”
tight, quotable verdict on the model's core weakness→ newsletter pull-quote↗ Tweet quote
43:30
“In under five minutes, it was able to index and review 222 pull requests. And this cost me a total of 10 cents on the contributor tier.”
concrete, surprising number that proves the value case→ TikTok hook↗ Tweet quote
The Script
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
17px
metaphoranalogystory
Something I'm a bit ashamed to admit. I'm a pretty big fan of Meta. Not of things like Facebook, like, no, gross, not my thing at all.
But when it comes to their actual software contributions and things they do for the ecosystem, I found them to be pretty solid, if not great. Projects like React and React Native were essential to the growth of the web and to an extent mobile as well and are genuinely incredible things that they've put out there for free for people to use however they want.
Their contributions to the AR and VR ecosystem were all things that I enjoyed heavily at the time because I'm a big VR nerd. And when they started getting into AI, they did that similarly well with their focus on open weight models with the LAMA line. To this day, LAMA is still used almost like a generic term for open weight models, and it's crazy how far the LAMA models went, but it's even crazier how far the entire industry has moved ahead of Meta.
They've been trying to correct course for a while with this hidden quietly worked on line of models called Muse. And historically, they have not been accessible beyond their little web interface to try it out. But apparently, they've been trying to make it ready for code, not just the model, but the tools around it as well.
And now, they're finally ready to release. Zuckerberg just announced on Twitter, by the way, that Muse code is out in beta today. This is their clone of Claude code, and it's powered by Muse Spark 1.2, which is their new coding focused model.
It's been a while since I've seen a lab publish benchmark numbers where they're not in first place in anything, but that doesn't mean it's a bad model. As you guys probably saw in my video about Grok 4.5, there is still a lot of value in well priced, fast, useful models.
And with Meta's focus on using every model for everything lately and collecting all the data they get as a result, it's actually gonna be pretty interesting to see how well this performs. So I'm gonna go through this with y'all together. I'm gonna try it live, I'm going to read through what they have to say, I'm gonna see what others are doing with it, get a feel for the price and the weird quirks as well.
And I just had to sign in to my terminal with Facebook, so I'm feeling a little gross. This is gonna be a journey, and I hope you enjoy it with me.
But first, a quick break for today's sponsor. There's a lot of code review bots around nowadays, and they're pretty good at reading the code. But I've noticed that running ClaudeCode and Codex on my machine tends to be better simply because the models can actually run things and verify the changes.
Wouldn't it be great if those review bots had the ability to actually verify changes on a real computer? GREPTILE thought so and that's why they introduced T Rex. It's a new sandbox that can actually run your code and verify the changes rather than just reading the syntax and hoping it's good.
If you've ever added an element to your UI and had it overlap something that was obvious when you ran it that the code review bots missed, that's what this is for. Because the models can actually click through and test things, and it can even respond with screenshots of the things it finds along the way. Their website lists a bunch of real world open source projects that have had bugs prevented through T Rex.
Even companies like WorkOS are finding real bugs. For example, this code that seems like a totally safe filter, but has a really rough edge case that you'll only find when you run it on real data. If you've shipped a bug in the last three months, you owe it to your team to check out soydev.link/reptile.
So let's start with what Zuckerberg had to say about their most recent release. Releasing Muse code in beta today. It's a terminal coding agent that takes on complex software engineering tasks across large repos, planning changes, writing code, validating the results.
It's powered by new Spark 1.2, which is a coding focused model update. This is gonna be an interesting one, because a lot of the other legacy companies like, I don't know, Microsoft and Google, don't necessarily have the best stack internally.
Whereas Facebook has done some incredible things with their internal tech. They didn't use Git because it was too slow for the scale they were moving at, so they built their own custom everything on top of Mercurial. And it's really powerful, to the point where people have been trying to copy their workflows with things like stacked PRs and stacked diffs.
I know a lot of people who used to work at Meta and left, and just missed all of that tooling that made it so much easier to work in these gigantic projects that are meta scale. Other companies like Google also have gigantic mono repos, but they are less focused on fixing them. In fact, they've went as far as laying off the teams that built the tooling to make them usable.
Meta's always done the opposite, where they'll rewrite a language if they have to, like they rewrote PHP and Hack, in order to try and make their code bases work better. Since Meta's code bases are so huge, it's often easier to rewrite the language that's underneath them than it is to rewrite the code base itself on a better language.
And that type of thinking is, at least it was, unique to Meta for a long time. And I think it kind of positions them well to dive into this world of using models and tools to work with your code bases at that scale. And that's why they've been using models like Fable and Opus so heavily recently, so it's interesting to see what they're cooking here.
They shared some benches here, and the first notable thing is that they're not putting Fable or Soul in these lists. They're only putting Opus and 5.6 Terra in, which is a choice. In Terminal Bench 2.1, they slightly beat out Terra and are slightly behind Opus five.
Those are a big jump from Use Spark 1.1 to 1.2. 1.1 was barely available to the public. 1.2 is now actually out, and the pricing's really interesting, which we'll show in a little bit.
DeepSwee, which is one of the benchmarks I prefer, it did pretty well here, getting a slightly higher score than Grok 4.5, which is nuts, because I actually found Grok 4.5 to be a very pleasant to use model, but it's still lagging behind Terra and Opus, which means it's far behind models like Soul and Fable.
And then Meta's internal coding bench, it came in second place in, with Muse Spark being much closer in third. And if this model is as far off as other benches showed, I think their internal bench might not be the best. They also, in their internal bench, have Terra and Gemini three six Flash Suspiciously close.
So we shall see as we go. Oh, god. They're already getting community noted.
Community note, Meta is still dead last in AI with this release. Okay, maybe not. Apple is, but they're behind just as much as Google Gemini is.
That's not a thing you need to put in a community note, nerds. Anyways, according to Zuck, Muse code runs specialized background agents that stay active your whole session, so they build up context over time instead of starting from scratch on every task. Interesting.
When a job is big enough, it fans out the separate sub agents working in parallel and isolated work trees. Your working copy is never touched. In testing, we had it build six features for a game simultaneously with no collisions.
Very interesting. Again, big code based stuff.
We pointed Muse Spark 1.2 at a kernel authorization task and let it run. A thousand tool calls over twenty four hours on Nvidia Hopper. It kept finding substantial improvements well beyond the initial exploration phase.
This is also kind of funny to me because Meta rented a bunch of TPUs from Google. They're one of the few companies to actually buy and rack hardware from Google. And And then Google went and rented a bunch of GPUs from Nvidia, all from xAI and SpaceX.
So Google is reselling their shitty chips to Meta, so they can use the money to go buy better chips from Nvidia through xAI. And now apparently, Zuckerberg is still stuck on Nvidia. Auditable by design, every model call, tool run, and edit has a local event log before it executes.
If it crashes mid task, it picks up exactly where it left off from that log. No lost work and no re prompting. Pricing.
It's easy and low cost to get started. Install Muse code with one line, and you can start on our contributor tier. Contributor tier is an interesting piece we'll talk about in a second.
Muse Spark one two is our next step as we push towards frontier with larger, more capable models on the way. Install it, use it, and tell us what you think. Well, the first thing I think is that when you Google search Muse Spark, you get all of these old articles about 1.1, this old developer page.
Okay, this one has 1.2 on it now, but it was actually kind of hard and annoying to find this initially. Here's the part I was looking for though, the price. Dollar 25 per mil in, 15¢ for cashed one mil in, and $4.25 per mil out.
Not great pricing until you realize I was hiding something from you. The contributor price tier.
This tier is 10¢ per mil in into 20¢ per mil out with point 2¢ per cashed input token. That is a 10 to 20 x price gap.
They're effectively giving out the contributor version for free because they need training data so goddamn badly. Meta's released Vue Spark 1.2. It's their third release in four months, and it scores a 54 on the intelligence index, significantly improving agentic knowledge work capabilities over prior releases, putting Meta next to SpaceX AI in a tie for third place amongst US labs.
Crazy they have to specify this now because the Chinese labs have caught up so much, especially with k three. Muse Spark 1.2 lands at a 54, up three points from 1.1, and 11 points from Muse Spark one point o, which came out in April.
That's a pretty big jump. Like, if we're just looking at the trajectory of what meta right here, that's a lot of improvement and not a whole lot of time. It's effectively tied with GPT 5.5 and Grok 4.5, narrowly behind current frontier models like Opus five, Fable five, and five point six oul, as well as Kimi k three.
It's among the most cost efficient models at its intelligence level. It's 40¢ per intelligence index task at Meta's unchanged dollar 25 and $4.25 per million token pricing.
They don't mention I I was right when I was reading that earlier. They are not measuring the prices based on the contributor tier. They are measuring them based on the normal pricing that they would charge actually.
So if we go back to the pricing chart here, it's about a 10 to 20 x difference in price, which means Muse Spark 1.2 on the contributor tier is actually the cheapest model currently here, because it would be 2 to 3¢ per task, which is comparable to v four Flash, as well as five six Luna.
Very interesting. This model, especially if you're down to, like, let Meta have your data, could be a really good value for the short term. A, Ab Initio's abstention rates have increased.
The model's score rose from an 18 to a 22, and the hallucination rate fell 10 points. The attempt rates dropped from 82 to 67. This is this bench is measuring how likely is the model to lie when it doesn't know something, and it's much less likely now.
Even though from my experience, that is not the case. Scientific reasoning results remain largely unchanged. Does much better at GDP val, but I still don't love that bench.
Good analysis. We can work with that. Muse is trained heavily on data from people using models like Anthropic Models internally, and all of the work they do.
I've heard numbers as high as 50% of staff at Meta have been moved over to some type of data labeling or like management tasks in order to help them train smarter models. And one of the results here is this wonderful tweet from Luke. Muse Spark is trained on the screen recordings of Meta employees, which makes it best in class at applying for jobs at Anthropic.
Absolute banger. I wanna go through other numbers that matter before showing off the code that I can write with Muse. I already have it working in the background, and it's surprisingly fast.
We'll talk about that in a sec. But first, I wanna look at the numbers from artificial analysis. As you can see here, Muse Spark 1.2 falls right between Kimi k three and Grok 4.5.
I will say it's a little bit embarrassing to release a closed weight model as a lab that has billions of dollars like Meta does, and have it be behind an open weight model that came out much before it. That is a little embarrassing, but also if it's really fast and cheap and reliable, these numbers don't mean everything.
Because I can tell you confidently that Opus five, despite being at the front of this list and tricking me into thinking it was a great model, ended up all being Copus because I hate using Opus five now. It Once I started merging the code it wrote, I realized how bad it was because I had to have Fable and Five Six Soul come in and clean up the mess.
Not great. But again, Kimi k three actually felt quite usable as did Grok 4.5, but Opus at the front doesn't feel usable for real code work at all.
But what I really wanna see is output tokens. How did they do here? Interesting.
Okay. So MuseSpark's runs did about 30 k tokens per task compared to about 36 for Fable five and compared to about 17 k for five six SOL.
So it's not quite as inefficient as Anthropic models, but it is far from the efficiency that we see from models like the GPT line or Grok 4.5. Very interesting. Definitely puts it in a weird different spot.
Like this model isn't just another open weight model tuned or something. Like this really is a new model, and it should feel quite different, especially when you look at the speed that this model has too. I don't think they have recent speed numbers yet on artificial analysis, but they do have them on OpenRouter.
OpenRouter is an easy way to use pretty much every model, and they track the throughput and speeds that users are actually seeing when they use the models through OpenRouter. And they're seeing an average of 191 tokens per second, which is absolutely nuts.
For reference, $5.06 SOL is only at about 30 tokens per second, so that is a gigantic improvement, like five x plus.
Absolute best case with SOL is one thirty four TPS, but realistically speaking, the p fifty's in the forties to fifties on OpenAI's official infra, and the fifties for Azure at best. To compare, Muse Spark gets 316 TPS at its best, and one sixty two at its p 50.
That is pretty nuts. And I feel the result coding with it already.
Man, that Cerebrus deployment of five six oul can't come fast enough. It was supposed to already be out, so I'm not sure what happened there. No one's given me any info, sadly.
But then we get to cost. And Muse Spark is performing absurdly well here at 40¢ per task on average through the Artificial Analysis Intelligence Index.
That's still more expensive than v four Flash is, as well as five six Luna on max, but it's a lot cheaper than other things. Even Mistral Medium is more expensive, and Kimi k three is more than twice the cost for similar work.
And I'm assuming this is the normal cost too, not that contributor tier, is effectively free. I think it's about time we fire off some prompts. I just opened Muse code inside of the t three code code base.
It's defaulting to one two contributor on high, and since t three code is fully open source, I don't mind if they get my data. They probably have it anyways. I asked it to give me an overview of the architecture of the code base, and it had this under thirty seconds.
Not bad. Let's see where this goes if I ask it for something a little harder, where I ask it to investigate more. I asked it if it can find anything suspicious about the event sourcing model, things cause weird behaviors on different platforms, unnecessary bandwidth usage and lag, etcetera.
Let's see what it does for that. Another interesting thing I noticed when I set this up is that it pulled all of my skills and personal rules from Claude code. Normally, things don't use the .claude directory because it's Claude's.
They use the shared dot agents standard and agent m d, stuff like that. They seem to be really focused on Claude. And from everything I know about how Meta's been working lately, they're definitely an anthropic house.
They use a lot of Opus and a lot of Fable there. Yeah, one seventy four TPS. Okay.
Okay. Now we're talking. For comparison, Grok 4.5, which is still quite fast, is at 50 to 52 TPS.
And a model like five six Soul is 30 to 50. It's crazy that Azure is performing that much better than OpenAI's official servers are right now, but yeah, you're welcome. I had to fight hard to get Azure to move.
It has some interesting findings here. I'm gonna tell it to summarize them. Alright.
So our findings in an easy to digest HTML file. Let's see if it's able to find my HTML plan skill and actually take all of this work and synthesize it. Look at that, already found the skill.
Now it is thinking, hopefully it will synthesize this properly and then put things up. It's interesting to state how much data this skill is. There's definitely some slop in here.
Things like the way that the full path for files is included in the UI and stuff, like, this CLI is a bit slop. Here is the HTML it made for me to summarize the findings that it had. Pure decider plus single serialized worker is correct, cost surfaces at the wire, full snapshots on subscribe, unbounded payloads, and a global sequence that couples every thread.
So here's what we'll do for a comparison. I'm gonna do a Claude run with Fable. We're gonna paste this.
I have a plan here. You know what? I wanna try DeepSeek four model, DeepSeek v four flash.
Let's copy over the same prompt then.
God, that is fast as fuck too. For reference, Muse 1.2 got me this feedback and even had an HTML page generate all in well under a minute. I've just been trying to get similar feedback from Fable, and we're over four minutes in and still don't have any useful anything here.
I also have DeepSeek Flash running in the background as well, just as a comparison, because the new Flash release was really solid too, just to get a comparison across things here. RG wasn't fine then because I used dash g quote star dot ts, which doesn't work on Mac OS with RipGrep. This is the DeepSeek one running, by the way.
God, these are both so slow and not getting me useful feedback so far in comparison. It's been very interesting. My plan is I'm gonna take the findings that Muse had and throw them at these other two models after and say, how do these compare?
One of my favorite ways to get a feel for what model strengths and weaknesses are is to have them all do the same task, have them write up the results, and then have them compare each other's results. DeepSeek Flash just finished its analysis. I'll ask it write up your findings in HTML, see if it can figure out my HTML skill.
Oh, use my HTML skill. You can read it from my dot agents or dot claud.
It doesn't have skills built in by default, if I recall. Oh, did, it found it nice. After waiting way too long for both Fable and DeepSeek to do a similar investigation, I asked them both to review their findings and compare it to what was found with Muse 1.2.
And here we see the way that Fable categorizes the differences, and remember, Fable, when doing a similar task with Opus, thought that Opus had a much better investigation overall. It seems like that is not the case with the Muse investigations. There are a few places where it said that Muse did better, for example, the coverage breadth it thinks Muse went a little further on.
But for the most part, it thinks it did worse and that Fable's plan was much more accurate and overall properly found root causes. Meanwhile, Deep Seek seems to think that the Mew's plan was meaningfully better.
Very interesting. Yeah, it's a solid model for the price, but I don't know if I would trust it for really heavy end to end work yet, but we'll test it a little more. I had to try one of my favorite tasks where I take my crappy fish game and have it remake it both in two d and in three d.
And the most impressive thing is how quickly it did it. It made the two d version about two and a half minutes and the three d version in under five. For reference, Opus five took over an hour for both of these.
It doesn't move properly. Like, I can't turn.
Kind of nuts that it made this whole thing from scratch in four minutes, though. Like, that part is impressive. It's just that the rest is broken.
Mouse look does not work. And it made a two d version of Fish Slop as well in two minutes and thirty nine seconds. For reference, when I put these same prompts to remake this game into other models like Opus, it took multiple hours to do.
And this made it in under five minutes for both the two d and three d versions. Is it perfect?
No. Far from it. But it does have a couple niceties.
Like it it made solid animations using the sprites that it had access to from the other project. It did shadows and like contrast well. Like it it has a weird type of taste that I haven't seen models have.
It's obviously still jank as shit in a lot of ways, but like the little shadows are nice, the way things are moving is nice, it feels fluid, like it's This model feels different. It doesn't feel like it's just distilled on other popular things.
It does have a vibe to the results, and I am definitely curious what's gonna happen when they make bigger and smarter versions. It allegedly just fixed the three d mouse broken stuff in a few seconds.
And it does appear that it has. I can now look around as expected.
The fish are all swimming backwards though. That's hilarious. Fish are swimming backwards.
I can't look up or down. Only left and right.
Can you use computer use to test it yourself? Oh, I found my codex computer use skill. Oh, no.
That is not the solution. As if it can do it itself without codex. Does it have vision?
Oh shit, it does. That's huge. Okay.
Now the up and down movement of the mouse tilts this. That is broken as shit.
And the fish are still moving backwards. It promised me it'd fix that. Oh, wait.
No, it didn't. Yes. Please fix and rebuild.
Okay. Nope. Fish moved the right way now, but I still can't look up and down.
Spaces both feed and go up. There's a lot of little things that got wrong here, like it it doesn't understand that things can collide with each other, which is interesting.
So one of the things they said the model was good at was like breaking things up and letting lots of stuff work at the same time. But it's it's getting work done, and it's doing it alarmingly fast.
Like, that's one of the nicest things here, is using it to like touch something up seems to be a useful method. Dara just got me one of my favorite things to look at when new models drop, the comparison of how it handles design on his witchai.dev site.
Very nice for, like, looking at the designs the different models do and comparing them. And I'm already seeing some interesting things here. Like, I love how it uses the scroll areas where this, like, graphic on the side scrolls with you until you hit a certain point.
The sections are tasteful, too. I I don't hate this. I'm tired of the pills on everything.
It feels a little template y, but not bad at all. And this little tilted reminder recall thing here is nice, too. This is all without the design skill as well.
If we switch to the second one, we end up with one of the cringey code style ones. It's laid out well. Like, its page layouts are solid so far, but I don't love the design and the graph it made there.
It really likes doing these tilted things. It has one here as well. I, actually, I really like this one.
Of them so far, this design looks decent, and I like the way it's using the, like, squircle here. I don't like that it switches from squircle to rounded there.
This isn't bad, though. This, again, is reminiscent of Gemini, to an extent.
Way too much text. Way too much text. But otherwise, not bad.
Ugh. These brutalist ones are getting tiring, I say, after making lawn video, which is absolutely of this style.
But, yeah, it's not bad. Like, nothing jumps out as, like, that is horrible.
I don't love having the bar here when it's already a very Brutalist, like, line heavy style. And now the classic, like, tailwind template homepage.
Oh, does these little floaties well. I I like how it uses those. Other models get that stuff really wrong, and I've fought it a ton on things like the t three code marketing site.
That's like what's interesting about it. This model just has a bit of a different flavor. You know what?
Let's give it something harder. Does it have work tree support, first off? It does.
Huge. That means I could play around with less risk a bit here. Can I yolo mode from here?
Nice. Oh, that's a bit annoying because I had that in a work tree. Dash dash work tree dash dash yolo.
Now we're in a safe work tree where I can ask it to do stupid things. Like, let's try their voice, actually. I would like to implement Muse as a provider inside of t three code.
I'm not sure what Muse code supports in terms of integration methods. I haven't played with it enough or looked into the SDK. I don't even know if it's open source.
What I really want is to integrate it through a layer like ACP, similar to how we have integrated other providers in the past, but if we need something more custom, I am down. I would like you to start by investigating everything you need here, both how we can implement within t three code, but also what is offered by Muse itself.
You'll have to do some digging to find source code where available, docs, SDKs, and whatnot. Might be a bit hard to find because this is all still so new.
Use a lot of sub agents to break up the work as you go. I actually liked that voice to text. It showed things as I spoke, which a lot of other solutions don't.
They just show it when you're done. So let's send that over and see how it goes. That's the first good terminal voice to text I've seen in any of these CLIs for being real.
Look at that. It's already spawning sub agents, and we can look at them almost the exact same way we can in Claude. I I know I clickbaited a bit saying this is a ClaudeCode clone, but like, this is such a Claude Code clone.
It's cool that it spun up all of these intelligently. Like, it's it's clear they are trying to make this model good at this type of, like, broken up, heavy sub agent breakup work.
I just got hit with a rate limit? Are you kidding? I put in a credit card and everything.
Are these sub agents getting hit with them too? They are. Great.
Why why would you make a sub agent flow like this if your APIs can't even handle it, Meta? Ugh. How much does it cost so far is a good question.
Interesting. I actually checked out the dashboard before. They have instructions on how to set it up with other providers too, like OpenCode, Cloud Code, Codex, as well as Curl and Python directly.
Assuming that Python is showing you how to like set it up with code, I'm assuming this is like an OpenAI compatible API. I wanna see how much I have spent. I don't know how accurate up to date this is, but everything I've done so far apparently is only about 13¢, including both of those game rewrites as well as the code base audit.
So, yeah. Absurdly cheap. It's making progress here.
Why is it talking about anti gravity? I just asked, why is so much of this about anti gravity? I didn't mention anti gravity at all.
It did a deep dive, it took under three minutes despite the fact that it was hitting rate limits. But, like, the results are useless. Contributor has very low rate limits.
Okay. That's the problem. It's because I'm on that contributor tier.
Okay. I shouldn't be so abused by this, but this is actually objectively hilarious. I asked why is so much of the response talking about anti gravity?
I didn't mention anti gravity at all. You're right. I assumed Muse equals anti gravity, Google's AGY, CLI, and anti gravity, because those names overlap heavily in recent search results, and I ran the whole investigation on that assumption without confirming.
I don't have a clear hit for Muse as a distinct product. Web searches mostly return Muse Spark, which is Meta's new coding model July 2026, an unreleased Muse Markdown OS project, something definitive for a Muse CLI or SDK. Can you point me to what you mean by Muse?
A URL, the company behind it, or what the binary or SDK is called would be enough. I'll give you a hint. I'm in it right now.
Let's see if this can figure out what it is. I'm way too amused. Yes.
Muse is Muse code itself, the CLI you're in right now, powered by Muse code o point one point o, powered by Muse Spark. My first investigation mapped it to anti gravity by mistake, completely wrong target.
So of course, the Muse Code layer here does not have a lot of the functionality that I would want to integrate with it. Things like it being open source, things like it having an SDK I can plug into ACP as a way to communicate with it, it's missing a lot of the stuff I need.
I do want to write up a plan on how we would integrate Muse code. You know what, I'm gonna send the smarter model through this with a smarter tool. T three code.
I'll do it on this machine so that I have the results here. I want to integrate the new MuseCode CLI from Meta into t three code as a new provider.
Not much documentation exists, but I do have it installed right now.
Investigate and help me figure out how we can integrate.
So I have Fable both investigating how it would do it, and separately in parallel, I have it reviewing the plan that was written by Muse. Oof, the page layout's super broken. Bit annoying to make those types of mistakes, though.
There's a lot of things here that feel really last generation, but a lot of things that feel next generation, too, where it's like, it seems to know how to break things up with sub agents really well, but it also seems to struggle meaningfully with, like, basic page layout stuff, or staying on task, and not hallucinating its way down an entirely incorrect path.
Do we get Omniscience scores here yet?
Apparently, it does well on AA Omniscience. That is weird to me, because we just watched it hallucinate aggressively. Allegedly, it is about as bad of hallucinations as Five Six Soul is and slightly better than Kimi k three, but considering how aggressively we just watched it do that, that was rough.
I had Fable give feedback on Muse's plan to integrate Muse, and it said it's a solid plan and gave a little bit of feedback. So I'm gonna do what I usually do in my real world day to day work when doing things like this, and I'm going to take that response, I'm going to paste it into the original chat with the first model that made the plan and see what it does.
Cool. It updated the plan alarmingly quickly as well. It's it's very nice how fast this model is.
It's it is fast and token efficient. It just feels great. Plan updated.
Take another look. It is nice having all of these fast, cheap models getting smart. Again, it's been a while because everybody's been fighting so hard on the frontier.
Considering that OpenAI just lowered the price of Luna by 80%, which is a massive drop on a model that was already really cheap, Dollar 20¢ per mil out and 20¢ per mil in is just unbelievable.
It's crazy that you can get a model with a million context window with the capabilities that Luna has for this price, and I've been using it a lot more for like random background tasks, categorization, titles, and stuff like that.
We were having almost no good small models drop for months, if not like a year, to having Luna come out and then price drop massively by like five x. DeepSeek v four Flash just got a new snapshot that I was testing earlier that is really good, really fast, and really cheap. And now we have this new Spark 1.2 model that's also seemingly really good for the money, especially if you're on that contributor tier where they are reading all of the things that get sent.
After the one pass of feedback from Fable, apparently it thinks that the plan Muse wrote is ready to go. That's kinda nuts. Man, I fucking love t three code.
It's so nice being able to just like hop between models, harnesses, and computers. I'm now spinning up Soul on one of my Linux boxes to compare the plans that Fable and Muse wrote separately.
See, we're up to 36¢ of spend now that we've done all of the planning. Oh, nope, 40¢.
Wow. I'm gonna go broke at this rate. For reference, just my little tests with Fable and Opus on this computer for reading those plans and investigating them is already $32 almost of spend in Claude.
So that is a 100 x gap, and I've done way less work with these models than I have with the Muse one. That is on the they have access to all my data tier, but to be fair, Fable five also does have them storing the data. They're not training on it, allegedly, but they are storing it for safety reasons.
Opus, they don't do that with. And if you spend the much higher 10 to 20 times more on the standard tier for Spark 1.2, then you will end up paying, still not anywhere near this much money, but if it's 10 to 20 times more, which is roughly what it is, that would be 4 to $8 at most.
For a bunch of work, that is decent. Oh, boy. It's spinning up the sub agents now.
Probably gonna hit rate limits again. Yeah, I'm hitting rate limits again. I might have to switch.
I'll stop all of these. We'll move off the contributor tier. Let's see how it handles continue.
It is nice only having one model and just picking between the two versions that have different pricing for the exact same thing. But when one of them hits rate limited that hard, it's a little annoying. I just spend way too much money in comparison, even though all the data I'm about to get here is technically available for anyone to train on, because it's in my videos.
The review sub agents are finishing up their work. I I switched to the higher price tier, and I'm still getting rate limited? Are you kidding?
Like, I'm just doing the sub agent stuff they said that they support well, and I'm hitting rate limits constantly. This is obnoxious. And it sucks, it's like the harness actually feels pretty good.
It's like a minimal polished up Claude code. It is making my laptop a little warmer than I would like, but yeah, this is I am annoyed about the rate limit thing more than anything else here.
The rest is not the worst. Like, rate limits on subscription plans make some sense. Rate limits on paid per token tiers, where I am using it the way it's intended, is pretty rough.
Oh, cool. Soul is now done comparing things here. Yeah, this is what I expected.
It's a slaughterhouse. It also decided to weight the different categories more and less heavily, depending on how it felt about them.
So it said for the current repository in API Fidelity that the plan for Muse is four out of 10, the plan from Fable is seven out of 10. For the protocol research, they tied roughly. For end to end and multi surface completeness, Muse was nowhere near there.
Fable was a lot further along at eight out of 10. Lifecycle, permission safety recovery, pretty close. Delivery plan, close ish, with Fable a little ahead at seven versus five.
Simplicity and maintainability, both didn't score great. But overall, Muse's plan got a 4.8 out of 10, and Fable's got a seven. Yeah.
Considering the gap in costs, that is reasonable, but considering the reality that you have to merge the code when it's done, this is much less reasonable. Still a very interesting model, just not necessarily one I would trust for making big sweeping changes to my code.
I think now is a good time to answer the important question, why would someone use this model? What makes Muse 1.2 useful enough that someone should use it?
Well, the obvious reason, like the number one thing that would make someone want to use this model is that they work at Meta. In which case, they would probably use this as their second or third most used model compared to Opus and Fable, because I know they love those models there. Main reason to use Muse is that you work at Meta.
But there are other reasons that I'm seeing a bit of. The biggest, of course, is that you like to try new things, especially when those things are cheap.
Because this model is cheap. Even if you're not willing to share your data and you're using it on the paid tier that's a little higher up, it's still a very, very cheap model. Although, now that I've switched out of that way cheaper tier, I just went from 40¢ to $5.32 for the work I've done in the past ten minutes, because it's going so fast.
But it did make its integration for Muse inside of t three code. So I will ask it to spin up dev server and share a link.
If it got this working properly first try after a little bit of plan review from Fable, then credit to them, they made a model that's pretty good. If this fails outright, then expected models this cheap can't really do tasks this exploratory and undefined, because again, they didn't put out the SDK I need to build this.
So the model had to integrate Muse into t three code with hacks more than anything. Moment of truth. Let's see if we are in here.
Muse Spark one two. What is this project? Not looking good so far.
Didn't even add it in the provider section in settings. It just added it in the UI here, and it does not appear to work.
Was a nice attempt. The auto generated title was from Codex, because that is the default. If you have Codex, I just use five six Luna on low for it.
So the title gen was nothing to do with Muse. It just appears to be broken. Okay.
So, sadly, what this means is you cannot trust it for longer running things, for sure. I was hoping it would be a little more capable at that, but it is not.
So what other reasons would you use this? You like really fast models that are somewhat capable.
Like, this is not anywhere near as good as a model like Five Five, even though the benches suggest otherwise. It just doesn't get it as well.
And that's what makes this model so strange to me, is that it understands breaking up work in sub agents and not stepping on each other's toes when it does that type of thing really well, but it's nowhere near as good at actually seeing complex work through. It's almost like it it it knows how to act like a modern smart model, but it doesn't know what the modern smart models know.
I like the comparison people are making with Grok Code fast, because it does feel similar there, where it has a lot of the the layers that make the models work well, but it's not good enough to really be trusted.
And I'll be honest, it's hard for me to justify using models like this for a lot of my work, just because I would rather wait two to three times longer and have something I can almost certainly merge, versus trying it five times with a fast model and still have a mess inside of it. Honestly, the most impressive thing here is the CLI.
It's a lot more stable and less annoying than Claude Code. It's still not my favorite, like, Py still smokes it overall, but it's solid for a thing that, like, they threw together for this release. I am tired of the labs making new CLIs and things when they're already in last place, forcing us to install yet another thing.
But they did provide integrations for things like OpenCode, so that was nice and right direction ish. And it still crushes everything Google is doing by quite a bit.
It is funny to see Meta quickly leapfrog Google, but still be far behind everyone else. It's almost like these big companies are fighting to be like fifth place, and all the startups that are really embracing the power of these new models are excelling and jumping far ahead. Obviously, Anthropic and OpenAI are far ahead.
Like, companies like Moonshot with Kimi, like ZAI with the GLM series and more are all far ahead of this in my opinion. Hell, I would still use Grok 4.5 above this.
I do have one last test I wanted to give it though. I want you to audit all of the pull requests I have open on t three code. Figure out which ones are mergeable, which ones need more work, which ones should be closed, which ones have been trumped by other things merging, etcetera.
I want you to make a priority list for me of what I should look at first, and how confident you are in me merging it. Break up this work into lots of sub agents in order to get through it faster. Cool.
We'll see how it does with that. Bad gateway when trying to hit the GitHub API through the CLI.
That's fun. Is GitHub down? No.
Is this formatting things wrong? Fun. It's doing some sketchy shit to get in.
Interesting. It went through all my PRs in four minutes. That's, like, genuinely impressive, especially because I spent the first two minutes just trying to get into GitHub and off it.
If this review is of decent quality, then this might be what I use the model for. I might just use this as my go to, like, go review all the PRs I have open.
Oh, that's a nice little thing. When I have Codex and Claude make these PR audit pages, they often don't link the actual PRs with these, like, links here, and this did.
That's really nice. I have been annoyed at how many of the models don't make these clickable links. Even gave little confidence scores.
This is one of the better PR review pages I've gotten. For reference, here's one I made with a different model, and I actually had to tell it, please make sure these are links. I genuinely prefer the layout of the version that Muse did here by quite a bit.
I might even use this as like my go to template in the future. This is super readable to me. I like little confidence scores.
I like that it tells you if it's clean or dirty merge, what it thinks the status of things are. This is good. This might be one of the use cases I end up using this for a bunch.
So to go back to use cases quick, video like fast models is one. One of the big ones I've now learned is for random code adjacent analysis work.
Obviously, wouldn't trust this model to, like, actually go merge PRs for me, but as a surface level, like go audit what's going on in this repo for 20 or 30¢, that was really good. Especially with like no additional effort to try and make it better.
Like that was great. I'm gonna tell it to go go to go further here. How about you do a similar audit for all open PRs that have had updates in the past five days?
If this audit costs less than like a dollar, it was absolutely worth it. One of the things that'll be hard for this model with this though is that it's not trained on GitHub, because again, Meta mostly uses their internal Mercurial stuff, where all the other labs are heavy on GitHub, so they've trained the model to be very good at GitHub.
I'm about to hit so many rate limits. It doesn't seem to have parallel limits like all the other harnesses do. It's more than happy to run seven sub agents in the background at once.
And it gave each of these a ton of PRs to look at. Those $8.41 before I hit send on this.
We have spent another 5¢ since. Yeah. It's gonna hit rate limits.
We'll let that run-in the background. I'll come back to it if it ends up with good results. But back to why you would use it.
I already said you like to try new things, but I really wanna emphasize this point, because this model has a different flavor. It like, when we went through the different designs it made, it did things meaningfully differently. It still has the like early Gemini three Pro era style overall to it.
Like a lot of these look like what I saw at a Gemini three and three point one Pro. But it also just has little niceties to it that give it a vibe that's different. It's it's like a little bit of seasoning that they added that other stuff doesn't have.
This really is a model for enthusiasts right now. Like, you wanna go play with it because it's fun to play with new things, not because this model is going to magically save you a bunch of money or become your go to, like no one should use this model as their day to day coding model. But what it is is interesting.
The things that Meta chose to focus on, the things they didn't focus on, and the capabilities that I'm seeing here, it's fascinating, genuinely. And I'm definitely gonna play with this model more. Probably through a better harness though, because you can't really integrate this one with anything.
I don't know why they closed source the Muse CLI here. Like, Muse code should just be open source. Meta, you guys know better.
You're an open source company at heart. Just put out the source. In under five minutes, it was able to index and review 222 pull requests.
And this cost me a total of 10¢ on the contributor tier. That is insane.
Being able to go through that much real work, like auditing 222 pull requests in my code base, organizing them by how mergeable it thinks they are, so I can quickly blast through this and ship real code. That's insane.
That's actually valuable. And this is why it's fun to experiment with the models, like try the different things you do and take a look at how much it costs and how fast it runs.
Being able to hit a button and spend 10¢, and in five minutes you have a page like this for 200 plus pull requests on your project, that's good.
That's useful. I'm impressed. I would use this regularly.
I might even set something up to automatically do this for me every day. That's cool as shit. Remember though, that was on the contributor tier, so if you're not willing to share this data to Meta, you're gonna be spending 20 ish times more, but that goes from 10¢ to $2.
For this type of work and this much work, that's genuinely really impressive. I I think you should play with this model if you're interested in this type of thing.
Obviously, you shouldn't trust it for everything. You shouldn't just blindly go through and merge stuff, but as a way to like pull signals out of noise for really, really cheap, it's a fun way to experiment and try things. Stuff like generating titles, categorizing PRs, generating summaries for reports, digging through logs to find useful stuff.
Like this is solid. And doing similar work with Fable costs tens if not hundreds of dollars, and this was literally 10¢. So yeah, not a bad model.
I think it's gonna get a lot of shit because it has all the weird quirks it has, but when you think about it for the price, the speed, and the capability, as well as its weird things it seems to do not better but different from other stuff, it's fun. It almost is like playing with one of those like interesting toy programming languages is how it feels to me.
It's it's different in a way that isn't necessarily ready for real world work all the time, but it's cool as fuck and very fun to play with. I definitely had ups and downs with this exploration, but overall I'm coming out somewhat impressed. I don't think this model's going to kill Opus or Fable anytime soon, but it's a model that I could see myself playing with for lots of weird things, especially when you consider the price.
If I didn't have all of these accounts across Claude Code and Codex that I just burn for all sorts of stuff, I would probably be leaning on models like this. And even though I have those other things I can burn, am still going to be trying this to try and just organize my work and life better because it is so surprisingly cheap.
I'm curious how y'all feel though. Am I too frontier pilled or is this model just not that impressive? Let me know how y'all feel and if you'll be using it.
Until next time, peace nerds.
The Hook
The bait, then the rug-pull.
Theo opens by admitting something he says he's a little ashamed of: he genuinely respects Meta's engineering, React, React Native, Llama, even as he can't stand Facebook itself. That respect is the setup for the real hook: Zuckerberg just announced Muse Code, Meta's own Claude Code clone, and Theo is about to find out live whether the company that builds great open-source tools can also build a coding agent worth using.
Frameworks
Named ideas worth stealing.
08:10concept
The Contributor Tier
A steep discount, 10-20x cheaper, that Meta offers on Muse Code pricing in exchange for letting Meta use submitted prompts and code as training data.
Steal forframing a free/data-sharing pricing tier against a private paid tier for any AI product
33:48list
Why Use Muse 1.2?
You work at Meta
You like to try new things, especially cheap ones
You like really fast models
For random code-adjacent analysis work: audits, categorization, summaries
The list Theo builds live on a whiteboard across the back half of the video, answering who this model is actually for.
Steal forwriting an honest 'who this is for' section in any tool review
Theo runs OpenAI's GPT-5.6-Sol through Claude Code instead of Codex and gets visibly better designs and cheaper orchestration — then reads Codex's system prompt on camera to find out why.
Theo spends a day stress-testing Moonshot's 2.8-trillion-parameter open-weight release — and comes away convinced it's frontier-class, cheap enough to matter, and genuinely dangerous once the weights go public on July 27.
Theo spends 36 minutes putting real numbers behind the GPT-5.6 hype — Sol, Terra, and Luna, benchmarked against Claude Fable, one blog chart at a time.
00:52
01:23
01:57
02:15
02:51
04:11
04:44
05:18
05:52
06:25
07:32
07:54
08:39
08:59
09:41
10:53
11:27
12:06
12:34
12:59
14:14
14:48
15:21
15:55
16:28
17:02
17:36
18:16
18:44
19:31
19:34
21:01
21:33
21:58
22:21
22:58
24:18
24:51
25:25
25:58
26:32
27:05
27:39
28:13
28:36
29:15
29:53
31:00
31:34
32:07
32:41
33:14
33:48
34:21
34:55
35:28
36:02
36:35
37:42
38:21
38:44
39:23
39:57
41:00
41:24
42:11
42:44
43:18
43:51
44:25
30:48
41:35
36:00
26:50
24:43
38:05