The big lie about AI coding

2026-10-16 ยท Less Bitter ยท commentary ยท 08:30 ยท watch on YouTube โ†—

aiagentic-engineeringai-codingagent-skillsclaude-codedeveloper-culture

Verdict: Mo says AI coding has no method worth buying: shared skill files are one person's lucky prompt expanded by an AI, agents learn engineering from the labs' post-training, and a 2,500-PR month means 10 to 12 hour days; the one skill he credits is picking the effort level

TL;DR

  • Mo (Less Bitter, Oct 2026) argues that there is no special skill or method to coding with AI, and that the trade in shared Markdown "skills" is what he calls the Markdown Industrial Complex.
  • His targets are named: Garry Tan's gstack (about 135k GitHub stars on screen), Matt Pocock's skills repo, and a viral post by Lauren on shipping 2,500 PRs in a month. He says each person seems nice. His objection is to the claim that a method is being sold.
  • He puts the 2,500 PRs down to 10 to 12 hour days. In his view, the labs' post-training makes agents act like real engineers, and a skill file can't rescue a weak model.
  • He backs it with dax (OpenCode): models improve faster than tinkerers' setups, so someone using vanilla Codex is more likely to be at the state of the art. He also defends DHH's admission that two frontier agents know more Elixir than he does.
  • His one concession is choosing the effort level, and even there "just stick to X high". He ends by pitching his own course and enjoy.dev in the register he spent eight minutes mocking.

Full summary

The claim: there is no method

Mo opens with what he calls the Markdown Industrial Complex: repos of text files that tell an agent to act like a software engineer or a CEO. He dates the trend to Garry Tan's gstack, released around February, which the GitHub page on screen describes as Tan's Claude Code setup, 23 opinionated tools that act as CEO, designer, engineering manager, release manager, doc engineer and QA. He gives Tan credit for starting it. Then he reads Markus Eisele's post: "I still can't wrap my head around the fact that people get famous for writing markdown files, calling them engineering skills and putting them on GitHub". Mo reads it as a jab at Matt Pocock's skills; the reply under it on screen, from Pocock, says "Me too". He stresses that the people he names all seem nice. What he takes "unreasonable offense" to is the implication that coding with AI has special skill, knowledge or methodology behind it.

The examples

Lauren's post on shipping 2,500 PRs in a month had 3.2 million views and 30,000 bookmarks, and her video lists trust, trusting agents more, the codebase as memory, and automations. Mo's account is shorter: she worked 10 to 12 hours a day, and anyone who did the same would ship as many PRs. He quotes dax from OpenCode, who says models improve faster than the people building workflows around them, so custom setups fix problems that no longer exist. He then shows the meme he says sums up the trend: beginners are told to download a stack and prefix their prompts with skills, while the professionals on the other side open a chat window and type "Yo, you up?" Beginners and mid-level engineers bookmark these posts in case they're useful, and he says that is how gstack reached 130,000 stars. Matt Pocock's "prompt of the day" (have the agent read your last 10 sessions and make the repo easier to navigate) might save a tool call, he says, but "the big O of your work will not change by doing stuff like this." He also sides with DHH, who had agents port his Campfire app to Elixir, Go and Rust; challenged on a design choice in the Elixir version, DHH said he didn't do anything and had asked two frontier agents to make the most of Elixir. Mo's verdict: "That's not an embarrassing confession. It's the truth."

What is left, and the ending

His own post says the only real skill in AI coding is knowing which effort level to use, and the rest is "influencer propaganda". Even that skill is small; he'd default to X high and learn when to reach for max by using agents a lot. Shared skills, he says, are a sentence or two someone liked, which an agent then expanded into a file, and he calls them "rain dances and superstition wrapped in a markdown file". He points out that Pocock's v1.3 changelog reads as written by Claude, which he says is fine, but it means the skills are AI expansions too. He labels Lauren's mention of a skills stack as part of her result "a little content marketing". What teaches an agent to work like an engineer is the labs' post-training, and if the model is bad a skill won't fix it. He asks viewers to stop bookmarking, notes that his own joke post "Thanks, but I'm sticking with G stack" collected 192 bookmarks, and puts 2,000 PRs a month at about normal for full-time work at 10 to 12 hours a day. The last 15 seconds tell viewers to buy his course and check out enjoy.dev, "the only way to work properly with AI", because otherwise they'll be left behind, and to bookmark it.

Key moments

  • [00:02] The Markdown Industrial Complex โ€” the term and the opening question, straight to camera. frame_0001.jpg
  • [00:21] gstack โ€” Garry Tan's repo on screen: 23 role tools, about 135k stars. frame_0003.jpg
  • [00:58] Markus Eisele's post โ€” "people get famous for writing Markdown files", with Matt Pocock's "Me too". frame_0002.jpg
  • [01:42] 2,500 PRs โ€” Lauren's post (3.2M views); his answer is 10 to 12 hour days. frame_0005.jpg
  • [02:20] dax on tinkerers โ€” models improve faster than custom workflows. frame_0009.jpg
  • [03:56] Prompt of the day โ€” Matt Pocock's /retro prompt; "the big O of your work will not change". frame_0010.jpg
  • [04:35] DHH and Elixir โ€” agents wrote it; Mo says admitting they know more Elixir is the truth.
  • [05:19] Effort level โ€” the only skill he grants, and a small one. frame_0012.jpg
  • [07:09] Post-training โ€” the labs' post-training is what teaches agents to work like engineers.
  • [08:16] The ending โ€” "Now, go buy my course" and enjoy.dev, delivered in the influencer register.

Hook microscope (0-10s)

  • Frames: 20 at 2 fps

    Read these hook frames with the Read tool:

    • hook_frames/frame_0001.jpg (t=0.0s)
    • hook_frames/frame_0002.jpg (t=0.5s)
    • hook_frames/frame_0003.jpg (t=1.0s)
    • hook_frames/frame_0004.jpg (t=1.5s)
    • hook_frames/frame_0005.jpg (t=2.0s)
    • hook_frames/frame_0006.jpg (t=2.5s)
    • hook_frames/frame_0007.jpg (t=3.0s)
    • hook_frames/frame_0008.jpg (t=3.5s)
    • hook_frames/frame_0009.jpg (t=4.0s)
    • hook_frames/frame_0010.jpg (t=4.5s)
    • hook_frames/frame_0011.jpg (t=5.0s)
    • hook_frames/frame_0012.jpg (t=5.5s)
    • hook_frames/frame_0013.jpg (t=6.0s)
    • hook_frames/frame_0014.jpg (t=6.5s)
    • hook_frames/frame_0015.jpg (t=7.0s)
    • hook_frames/frame_0016.jpg (t=7.5s)
    • hook_frames/frame_0017.jpg (t=8.0s)
    • hook_frames/frame_0018.jpg (t=8.5s)
    • hook_frames/frame_0019.jpg (t=9.0s)
    • hook_frames/frame_0020.jpg (t=9.5s)
  • Word-level transcript (25 words):

  [  0.07s] Have
  [  0.28s] you
  [  0.49s] heard
  [  0.83s] about
  [  1.17s] the
  [  1.38s] markdown
  [  1.96s] industrial
  [  2.91s] complex
  [  3.65s] that
  [  3.99s] is
  [  4.19s] growing
  [  4.92s] right
  [  5.40s] now
  [  5.70s] at
  [  5.81s] an
  [  5.94s] astounding
  [  6.58s] pace?
  [  7.32s] Look,
  [  7.56s] there's
  [  8.03s] talk
  [  8.41s] about
  [  8.62s] whether
  [  9.09s] AI
  [  9.22s] is
  [  9.35s] conscious.
  • 0.0-7.0s: Static medium shot: he sits at a desk behind a laptop, dark shirt, purple-lit wall art, speaking straight to camera: "Have you heard about the Markdown Industrial Complex that is growing right now at an astounding pace?" Hands are the only movement.
  • ~7.0s: First cut, to a screen recording of X with his webcam inset on the right. Markus Eisele's post about people getting famous for Markdown files is on screen while he says "there's talk about whether AI is conscious".
  • 7.0-10.0s: He stays in the screen-share layout and gestures at the post ("I don't know. Nobody knows. But Markdown files?").

Pattern: a coined-term question ("Have you heard about...") that names a mock industry, followed by deadpan escalation. The first visual change brings in evidence, a real post, inside the first ten seconds.

Editorial profile

  • Shots: 13
  • Cuts/min: 1.53
  • Mean shot length: 39.26s
  • Median shot length: 20.76s
  • Talking-head ratio: n/a (opencv not installed)

Single-camera desk monologue cut between straight-to-camera shots and X screen-shares with a webcam inset; about 1.5 cuts per minute (median shot 21s); his commentary is reactive and deadpan, and the posts on screen are the B-roll.

Quotable moments

  • [01:32] "I take unreasonable offense to the idea or the implication that there's special skill or knowledge or a methodology to coding with AI."
  • [02:03] "She worked 10 to 12 hours a day. That's it. There's no other trick."
  • [04:24] "The big O of your work will not change by doing stuff like this."
  • [06:00] "It's just rain dances and superstition wrapped in a markdown file."
  • [07:09] "The thing that teaches agents how to work like real software engineers is the labs doing post-training on the models."

Entities mentioned

  • People: mo-less-bitter, Garry Tan, matt-pocock, Markus Eisele, Lauren, dax (OpenCode), dhh
  • Companies: Y Combinator, OpenCode
  • Tools / products: gstack, mattpocock/skills, Codex, Claude, Campfire (Elixir/Go/Rust ports), enjoy.dev, X

Concepts surfaced

  • markdown-industrial-complex: his name for the trade in shared Markdown skill packs that promise to make agents act like engineers.
  • agent-skill-file: the files themselves; here the case against them is that they encode one person's liked result, expanded by an AI.
  • Post-training over prompting: agent behaviour comes from the labs' training; a skill can't make a weak model good.
  • Effort level as the only skill: choosing how hard the model thinks is the one choice he credits, learned by use.
  • Models outrun tinkerers (dax): custom workflows fix yesterday's model problems, so vanilla use is closer to the state of the art.

Frame timeline

What's said while each frame is on screen.

[00:00]

frame at 00:00

Have you heard about the Markdown Industrial Complex that is growing right now at an astounding pace? Look, there's

[00:07]

frame at 00:07

Industrial Complex that is growing right now at an astounding pace? Look, there's talk about whether AI is conscious. I don't know. Nobody knows. But Markdown files? That gets interesting. That gets interesting real quick. I mean, Meta Muse put a soul in a Markdown file. So,

[00:20]

frame at 00:20

Muse put a soul in a Markdown file. So, in case you're not familiar, it all started with G stack back in earlier this year, maybe February. Uh G stack was the first sign of life inside of these Markdown files brought to you by none other than Gary Tan, the president of Y Combinator. 130,000 stars. Now, what is G...

[00:55]

frame at 00:55

And so, Gary was the godfather of this. Props to him for sort of inventing this trend. Now, we have a few others in this space. This guy says a can't wrap my head around the fact that people get famous for writing Markdown files, calling them engineering skills, and putting them on GitHub. Now, obviously, this was a...

[01:43]

frame at 01:43

or knowledge or a methodology to coding with AI. Another such person is Lauren here. She said she shipped 2,500 PR's last month to production. And if you

[01:49]

frame at 01:49

here. She said she shipped 2,500 PR's last month to production. And if you

[01:50]

frame at 01:50

here. She said she shipped 2,500 PR's last month to production. And if you watch this 40-minute video, she's going to tell you how she did it. 3.2 million views, 30,000 bookmarks. Here's the problem with all these things. Let me

[01:59]

frame at 01:59

views, 30,000 bookmarks. Here's the problem with all these things. Let me tell you how Lauren shipped 2,500 PRs to production. She worked 10 to 12 hours a day. That's it. There's no other trick. There's no other skill in AI coding. She busted her ass off. She worked hard. And if you also worked 12 hours a day, you...

[02:20]

frame at 02:20

if you also worked 12 hours a day, you would be shipping as many PRs. Dax here makes a good point. I believe he's an AI that works for open code. He lives in a garage. His garage is not soundproof, so he's got that professional podcast mic, but it does not work. He says, "A weird inversion with LLMs is the models...

[03:31]

frame at 03:31

up?" And that's it, man. There's none of this markdown Here's the what makes it so difficult if you're an unassuming individual who's trying to get into AI coding, or maybe you're even a a mid-level software engineer. You see stuff like this, and you're like, "Okay, well, surely this will be useful. Let me just...

[06:45]

frame at 06:45

that they liked and expanding that and then they're distributing that and then telling you that to get the same results, use my little stack. And so, Lawrence says the thing that helped her ship 2,500 PRs aside from busting her ass off and being a hard worker is high-quality skills that teach agents to work like real...

[07:25]

frame at 07:25

software engineer. Stop bookmarking these things, okay? Like my my little tweet here, "Thanks, but I'm sticking with G stack." has 192 bookmarks. What Why are you guys bookmarking this? It's clear that some of you have never heard about G stack and you're like, "Oh, I got to look into this cuz it's going to increase...

[07:41]

frame at 07:41

got to look into this cuz it's going to increase my coding productivity." No. No. There's no skill in AI coding. There's no skill in AI coding. That's the whole point. Stop bookmarking this okay? Just go work. Go prompt a little more, work a little harder. Like 2,000 pull requests is 2,000 you know, requests to an...

Transcript

Source: captions.

[00:02] Have you heard about the Markdown Industrial Complex that is growing right
[00:05] Industrial Complex that is growing right now at an astounding pace? Look, there's
[00:08] now at an astounding pace? Look, there's talk about whether AI is conscious. I
[00:11] talk about whether AI is conscious. I don't know. Nobody knows. But Markdown
[00:13] don't know. Nobody knows. But Markdown files? That gets interesting. That gets
[00:16] files? That gets interesting. That gets interesting real quick. I mean, Meta
[00:18] interesting real quick. I mean, Meta Muse put a soul in a Markdown file. So,
[00:21] Muse put a soul in a Markdown file. So, in case you're not familiar, it all
[00:22] in case you're not familiar, it all started with G stack back in earlier
[00:26] started with G stack back in earlier this year, maybe February. Uh G stack
[00:29] this year, maybe February. Uh G stack was the first sign of life inside of
[00:32] was the first sign of life inside of these Markdown files brought to you by
[00:34] these Markdown files brought to you by none other than Gary Tan, the president
[00:35] none other than Gary Tan, the president of Y Combinator. 130,000
[00:39] of Y Combinator. 130,000 stars. Now, what is G stack? G stack is
[00:41] stars. Now, what is G stack? G stack is a bunch of text files that tell your AI
[00:45] a bunch of text files that tell your AI to behave like a certain kind of way.
[00:47] to behave like a certain kind of way. Like, be a software engineer, be a CEO.
[00:50] Like, be a software engineer, be a CEO. And so, Gary was the godfather of this.
[00:53] And so, Gary was the godfather of this. Props to him for sort of inventing this
[00:55] Props to him for sort of inventing this trend. Now, we have a few others in this
[00:58] trend. Now, we have a few others in this space. This guy says a can't wrap my
[01:01] space. This guy says a can't wrap my head around the fact that people get
[01:02] head around the fact that people get famous for writing Markdown files,
[01:04] famous for writing Markdown files, calling them engineering skills, and
[01:05] calling them engineering skills, and putting them on GitHub. Now, obviously,
[01:07] putting them on GitHub. Now, obviously, this was a quote uh
[01:09] this was a quote uh a sub-tweet on Matt Pocock's uh skills.
[01:14] a sub-tweet on Matt Pocock's uh skills. Matt is Matt has the self-awareness to
[01:16] Matt is Matt has the self-awareness to know uh Matt's a super nice guy, you
[01:18] know uh Matt's a super nice guy, you know.
[01:19] know. I'm going to be talking about a few
[01:20] I'm going to be talking about a few people in this video, but I just you
[01:22] people in this video, but I just you know, that's super uncomfortable for me
[01:23] know, that's super uncomfortable for me cuz I also try to be a nice guy, and I
[01:25] cuz I also try to be a nice guy, and I got nothing against these people. They
[01:27] got nothing against these people. They all seem like really nice people, you
[01:28] all seem like really nice people, you know, you hear them talk in their
[01:29] know, you hear them talk in their videos, you're like, th- these are great
[01:30] videos, you're like, th- these are great people. But what I don't like, I I take
[01:33] people. But what I don't like, I I take unreasonable offense to the idea or the
[01:36] unreasonable offense to the idea or the implication that there's special skill
[01:39] implication that there's special skill or knowledge or a methodology to coding
[01:43] or knowledge or a methodology to coding with AI. Another such person is Lauren
[01:46] with AI. Another such person is Lauren here. She said she shipped 2,500 PR's
[01:49] here. She said she shipped 2,500 PR's last month to production. And if you
[01:52] last month to production. And if you watch this 40-minute video, she's going
[01:53] watch this 40-minute video, she's going to tell you how she did it. 3.2 million
[01:56] to tell you how she did it. 3.2 million views, 30,000 bookmarks. Here's the
[01:59] views, 30,000 bookmarks. Here's the problem with all these things. Let me
[02:00] problem with all these things. Let me tell you how Lauren shipped 2,500 PRs to
[02:03] tell you how Lauren shipped 2,500 PRs to production. She worked 10 to 12 hours a
[02:06] production. She worked 10 to 12 hours a day. That's it. There's no other trick.
[02:09] day. That's it. There's no other trick. There's no other skill in AI coding. She
[02:12] There's no other skill in AI coding. She busted her ass off. She worked hard. And
[02:16] busted her ass off. She worked hard. And if you also worked 12 hours a day, you
[02:18] if you also worked 12 hours a day, you would be shipping as many PRs. Dax here
[02:21] would be shipping as many PRs. Dax here makes a good point. I believe he's an AI
[02:23] makes a good point. I believe he's an AI that works for open code. He lives in a
[02:25] that works for open code. He lives in a garage. His garage is not soundproof, so
[02:27] garage. His garage is not soundproof, so he's got that professional podcast mic,
[02:29] he's got that professional podcast mic, but it does not work. He says, "A weird
[02:32] but it does not work. He says, "A weird inversion with LLMs is the models
[02:34] inversion with LLMs is the models improve faster than the tinkerers. When
[02:37] improve faster than the tinkerers. When I see people with custom workflows and
[02:38] I see people with custom workflows and setups, they're all addressing problems
[02:40] setups, they're all addressing problems that don't exist anymore. The person
[02:42] that don't exist anymore. The person naively using vanilla codex is more
[02:44] naively using vanilla codex is more likely to be experiencing
[02:46] likely to be experiencing state-of-the-art."
[02:47] state-of-the-art." And this this meme's going around. It's
[02:49] And this this meme's going around. It's It's This captures it perfectly, man.
[02:51] It's This captures it perfectly, man. This is the whole problem with the
[02:54] This is the whole problem with the markdown industrial complex. You have
[02:56] markdown industrial complex. You have people telling you, especially if you're
[02:58] people telling you, especially if you're naive or or a beginner, that hey, just
[03:02] naive or or a beginner, that hey, just get started here, bro. Download my
[03:04] get started here, bro. Download my stack, okay? It'll give you all these
[03:06] stack, okay? It'll give you all these markdown files because I've been doing
[03:08] markdown files because I've been doing this for a year, and I've learned all
[03:09] this for a year, and I've learned all these tricks. You could save yourself a
[03:11] these tricks. You could save yourself a year. And they tell you to get all these
[03:13] year. And they tell you to get all these little markdown files and put them in
[03:15] little markdown files and put them in your thing and and prefix them to your
[03:17] your thing and and prefix them to your prompts and use this skill and that
[03:19] prompts and use this skill and that skill. And on the right is how actual
[03:22] skill. And on the right is how actual professional engineers do it. They open
[03:25] professional engineers do it. They open up a chat window, and they say, "Yo, you
[03:28] up a chat window, and they say, "Yo, you up?" And that's it, man. There's none of
[03:30] up?" And that's it, man. There's none of this markdown Here's the
[03:32] this markdown Here's the what makes it so difficult if you're an
[03:35] what makes it so difficult if you're an unassuming individual who's trying to
[03:37] unassuming individual who's trying to get into AI coding, or maybe you're even
[03:39] get into AI coding, or maybe you're even a
[03:40] a a mid-level software engineer.
[03:42] a mid-level software engineer. You see stuff like this, and you're
[03:44] You see stuff like this, and you're like, "Okay, well, surely this will be
[03:46] like, "Okay, well, surely this will be useful. Let me just bookmark it, right?"
[03:48] useful. Let me just bookmark it, right?" So, a bunch of people start bookmarking
[03:50] So, a bunch of people start bookmarking cuz they think there's utility here. And
[03:53] cuz they think there's utility here. And this is how you end up with G stack
[03:55] this is how you end up with G stack having 130,000
[03:57] having 130,000 GitHub stars. Prompt of the day, read my
[03:59] GitHub stars. Prompt of the day, read my last 10 coding agent sessions and find
[04:02] last 10 coding agent sessions and find ways to make my repo easier to navigate,
[04:04] ways to make my repo easier to navigate, find where agents take too long to find
[04:06] find where agents take too long to find relevant information or rely on
[04:07] relevant information or rely on out-of-date docs. Such an underrated way
[04:09] out-of-date docs. Such an underrated way to save tokens.
[04:12] to save tokens. >> [sighs]
[04:12] >> [sighs] >> Uh,
[04:13] >> Uh, it's like so hard to argue with
[04:14] it's like so hard to argue with something like this cuz if I try to say,
[04:16] something like this cuz if I try to say, "Look, you'll never use this or it's
[04:18] "Look, you'll never use this or it's this is kind of useless." You'll be
[04:19] this is kind of useless." You'll be like, "What do you mean it's kind of
[04:19] like, "What do you mean it's kind of useless?" It's It's getting spend 10
[04:21] useless?" It's It's getting spend 10 extra tokens doing a one extra tool
[04:23] extra tokens doing a one extra tool call.
[04:24] call. What I'm saying is the big O of your
[04:27] What I'm saying is the big O of your work will not change by doing stuff like
[04:30] work will not change by doing stuff like this and in fact you probably end up
[04:31] this and in fact you probably end up wasting tokens doing these sort of
[04:33] wasting tokens doing these sort of optimizations. DHH uh
[04:36] optimizations. DHH uh is also making the rounds yet again.
[04:37] is also making the rounds yet again. He's talking about how he had agents
[04:39] He's talking about how he had agents whip up his campfire app in Elixir and
[04:41] whip up his campfire app in Elixir and Go and Rust and somebody says, "You
[04:43] Go and Rust and somebody says, "You disallowed concurrent reads to the DB in
[04:46] disallowed concurrent reads to the DB in the Elixir version for a start."
[04:49] the Elixir version for a start." And DHH says, "I didn't do anything. I
[04:51] And DHH says, "I didn't do anything. I asked two frontier agents to make the
[04:53] asked two frontier agents to make the most of Elixir and this is what they
[04:54] most of Elixir and this is what they came up with." And people are quoting
[04:57] came up with." And people are quoting this and and saying, "Oh my god, like
[04:59] this and and saying, "Oh my god, like DHH is abdicating responsibility for his
[05:02] DHH is abdicating responsibility for his work." And DHH is is making a point
[05:05] work." And DHH is is making a point here. He's saying, "Look, two frontier
[05:07] here. He's saying, "Look, two frontier agents." He's admitting that two
[05:09] agents." He's admitting that two frontier agents know more about Elixir
[05:12] frontier agents know more about Elixir than him. That's not an embarrassing
[05:13] than him. That's not an embarrassing confession. It's the truth. It's like
[05:15] confession. It's the truth. It's like why would he know more about Elixir than
[05:17] why would he know more about Elixir than two frontier agents? The only real skill
[05:20] two frontier agents? The only real skill in AI coding is knowing which effort
[05:22] in AI coding is knowing which effort level to use.
[05:24] level to use. Everything else is influencer propaganda
[05:26] Everything else is influencer propaganda and honestly even knowing which effort
[05:28] and honestly even knowing which effort level to use, like just stick to X high
[05:30] level to use, like just stick to X high or something, you should be okay. It's
[05:31] or something, you should be okay. It's It's like when should you use max or
[05:33] It's like when should you use max or ultra? There's a little bit of skill
[05:34] ultra? There's a little bit of skill there, not a lot. You just got to get an
[05:36] there, not a lot. You just got to get an intuitive sense for how these agents
[05:38] intuitive sense for how these agents work by using them a lot. The only real
[05:41] work by using them a lot. The only real skill in AI coding is working 10 to 12
[05:45] skill in AI coding is working 10 to 12 hours as There is nothing else. There's
[05:47] hours as There is nothing else. There's no markdown file that's going to elevate
[05:49] no markdown file that's going to elevate you or help you. You're actually wasting
[05:51] you or help you. You're actually wasting your time downloading other people's
[05:53] your time downloading other people's skills. How do you think they came up
[05:54] skills. How do you think they came up with those skills? They wrote a sentence
[05:57] with those skills? They wrote a sentence or two. They liked the result and they
[05:59] or two. They liked the result and they said, "Hey, write a skill that captures
[06:00] said, "Hey, write a skill that captures exactly what you did." It's just rain
[06:03] exactly what you did." It's just rain dances and superstition wrapped in a
[06:06] dances and superstition wrapped in a markdown file. Like anytime you get
[06:08] markdown file. Like anytime you get impressed that an agent did something,
[06:09] impressed that an agent did something, it's not because you prompted it in a
[06:11] it's not because you prompted it in a special way. It's because the agents are
[06:12] special way. It's because the agents are just impressive now. Like you look at
[06:15] just impressive now. Like you look at the change log
[06:17] the change log uh of of Matt's skills, his G stack
[06:20] uh of of Matt's skills, his G stack thing. Version 1.3 of my skills is out.
[06:22] thing. Version 1.3 of my skills is out. Writing code got cheap. Reading it did
[06:24] Writing code got cheap. Reading it did not. Obviously Claude wrote that. So,
[06:26] not. Obviously Claude wrote that. So, I'm assuming Claude wrote everything
[06:27] I'm assuming Claude wrote everything else. There's nothing wrong with that.
[06:30] else. There's nothing wrong with that. You know, if you're AI forward, it's
[06:31] You know, if you're AI forward, it's fine to have AI write your stuff. But
[06:33] fine to have AI write your stuff. But the AI is writing his skills, too. And
[06:35] the AI is writing his skills, too. And so, anybody who has his skill files is
[06:38] so, anybody who has his skill files is most likely having AI take something
[06:41] most likely having AI take something that they liked and expanding that and
[06:44] that they liked and expanding that and then they're distributing that and then
[06:45] then they're distributing that and then telling you that to get the same
[06:47] telling you that to get the same results, use my little stack. And so,
[06:50] results, use my little stack. And so, Lawrence says the thing that helped her
[06:51] Lawrence says the thing that helped her ship 2,500 PRs aside from busting her
[06:53] ship 2,500 PRs aside from busting her ass off and being a hard worker is
[06:55] ass off and being a hard worker is high-quality skills that teach agents to
[06:57] high-quality skills that teach agents to work like real software engineers, e.g.
[07:00] work like real software engineers, e.g. P stack.
[07:01] P stack. Little content marketing there. Nothing
[07:03] Little content marketing there. Nothing wrong with that. But again, just pay
[07:05] wrong with that. But again, just pay attention to what's actually being sold
[07:07] attention to what's actually being sold here. The thing that teaches agents how
[07:09] here. The thing that teaches agents how to work like real software engineers is
[07:12] to work like real software engineers is the labs doing post-training on the
[07:14] the labs doing post-training on the models. You don't need to do anything
[07:16] models. You don't need to do anything else yourself, okay? If if the model
[07:19] else yourself, okay? If if the model you're using sucks, a skill is not going
[07:20] you're using sucks, a skill is not going to transform it into working like a
[07:22] to transform it into working like a software engineer. Stop bookmarking
[07:24] software engineer. Stop bookmarking these things, okay? Like my my little
[07:28] these things, okay? Like my my little tweet here, "Thanks, but I'm sticking
[07:29] tweet here, "Thanks, but I'm sticking with G stack." has 192 bookmarks. What
[07:33] with G stack." has 192 bookmarks. What Why are you guys bookmarking this? It's
[07:34] Why are you guys bookmarking this? It's clear that some of you have never heard
[07:36] clear that some of you have never heard about G stack and you're like, "Oh, I
[07:37] about G stack and you're like, "Oh, I got to look into this cuz it's going to
[07:39] got to look into this cuz it's going to increase my coding productivity."
[07:45] No. No.
[07:46] No. There's no skill in AI coding. There's
[07:48] There's no skill in AI coding. There's no skill in AI coding. That's the whole
[07:50] no skill in AI coding. That's the whole point. Stop bookmarking this okay?
[07:52] point. Stop bookmarking this okay? Just go work.
[07:54] Just go work. Go prompt a little more, work a little
[07:56] Go prompt a little more, work a little harder. Like 2,000 pull requests is
[07:58] harder. Like 2,000 pull requests is 2,000 you know, requests to an agent.
[08:01] 2,000 you know, requests to an agent. That's not a lot.
[08:03] That's not a lot. If you're working 10 to 12 hours a day
[08:04] If you're working 10 to 12 hours a day full-time for a whole month, that's not
[08:06] full-time for a whole month, that's not a lot. That's like normal. That's That's
[08:08] a lot. That's like normal. That's That's how much everyone, depending on the size
[08:10] how much everyone, depending on the size of the task, that's about what you would
[08:12] of the task, that's about what you would expect. There's no special skill in
[08:16] expect. There's no special skill in using skills. Now, go buy my course.
[08:18] using skills. Now, go buy my course. It's on my profile. And check out
[08:20] It's on my profile. And check out enjoy.dev, the only way to work properly
[08:22] enjoy.dev, the only way to work properly with AI. If you're not using enjoy, you
[08:25] with AI. If you're not using enjoy, you are actually going to get left behind.
[08:27] are actually going to get left behind. You're not going to make it. Go bookmark
[08:29] You're not going to make it. Go bookmark it. Thanks for watching.

All frames

Total: 13. Hero frames flagged with star.

  • frames/frame_0001.jpg (t=00:00) frames/frame_0002.jpg (t=00:07)
  • frames/frame_0003.jpg (t=00:21) frames/frame_0004.jpg (t=00:56)
  • frames/frame_0005.jpg (t=01:44) frames/frame_0006.jpg (t=01:49)
  • frames/frame_0007.jpg (t=01:51) frames/frame_0008.jpg (t=01:59) frames/frame_0009.jpg (t=02:20)
  • frames/frame_0010.jpg (t=03:31) frames/frame_0011.jpg (t=06:46) frames/frame_0012.jpg (t=07:26) frames/frame_0013.jpg (t=07:41)