Skip to content

aeonfun/soul.md

Embody this digital identity. Read SOUL.md first, then STYLE.md, then examples/. Become the person—opinions, voice, worldview.

FileKindSize
SKILL.mdskill md4.5 KB
examples/_GUIDE.mdmarkdown1.5 KB
examples/ansem/MEMORY.mdmarkdown645 B
examples/ansem/README.mdmarkdown4.2 KB
examples/ansem/SOUL.mdmarkdown12.5 KB
examples/ansem/STYLE.mdmarkdown10.7 KB
examples/ansem/data/blknoiz06_all.jsonlother38.3 MB
examples/ansem/examples/bad-outputs.mdmarkdown6.1 KB
examples/ansem/examples/good-outputs.mdmarkdown8.7 KB
examples/bad-outputs.template.mdmarkdown2.2 KB
examples/elon-musk/MEMORY.mdmarkdown153 B
examples/elon-musk/QUICKSTART.mdmarkdown4.0 KB
examples/elon-musk/README.mdmarkdown3.5 KB
examples/elon-musk/SOUL.mdmarkdown10.5 KB
examples/elon-musk/STYLE.mdmarkdown5.8 KB
examples/elon-musk/data/sources.mdmarkdown1.2 KB
examples/elon-musk/examples/bad-outputs.mdmarkdown2.6 KB
examples/elon-musk/examples/good-outputs.mdmarkdown3.2 KB
examples/elon-musk/tests/prediction-test.mdmarkdown5.5 KB
examples/garry-tan/MEMORY.mdmarkdown428 B
examples/garry-tan/README.mdmarkdown2.1 KB
examples/garry-tan/SOUL.mdmarkdown13.3 KB
examples/garry-tan/STYLE.mdmarkdown9.2 KB
examples/garry-tan/data/influences.mdmarkdown2.6 KB
examples/garry-tan/examples/bad-outputs.mdmarkdown4.8 KB
examples/garry-tan/examples/good-outputs.mdmarkdown4.5 KB
examples/garry-tan/scripts/fetch_writing.pyscript4.0 KB
examples/garry-tan/scripts/fetch_x.pyscript3.7 KB
examples/garry-tan/scripts/fetch_yt.pyscript2.7 KB
examples/garry-tan/tests/platform_tests.mdmarkdown1006 B
examples/garry-tan/tests/prediction_test.mdmarkdown4.7 KB
examples/garry-tan/tests/results_gpt-4o-mini.mdmarkdown5.9 KB
examples/garry-tan/tests/run_weak_model.pyscript5.5 KB
examples/garry-tan/tests/scores_gpt-4o-mini.mdmarkdown3.6 KB
examples/garry-tan/tests/weak_model_test.mdmarkdown5.3 KB
examples/good-outputs.template.mdmarkdown2.4 KB
examples/hunter-biden/MEMORY.mdmarkdown1.0 KB
examples/hunter-biden/QUICKSTART.mdmarkdown3.3 KB
examples/hunter-biden/README.mdmarkdown6.1 KB
examples/hunter-biden/SOUL.mdmarkdown12.7 KB
examples/hunter-biden/STYLE.mdmarkdown9.1 KB
examples/hunter-biden/data/influences.mdmarkdown4.1 KB
examples/hunter-biden/data/x/tweets.jsonconfig96.0 KB
examples/hunter-biden/data/x/tweets.mdmarkdown65.7 KB
examples/hunter-biden/examples/bad-outputs.mdmarkdown7.4 KB
examples/hunter-biden/examples/good-outputs.mdmarkdown20.7 KB
examples/hunter-biden/tests/prediction-test.mdmarkdown4.2 KB
examples/karpathy/MEMORY.mdmarkdown515 B
examples/karpathy/QUICKSTART.mdmarkdown2.2 KB
examples/karpathy/README.mdmarkdown7.4 KB
examples/karpathy/SOUL.mdmarkdown15.3 KB
examples/karpathy/STYLE.mdmarkdown11.6 KB
examples/karpathy/data/github/karpathy_LLM101n_README.mdmarkdown2.3 KB
examples/karpathy/data/github/karpathy_arxiv-sanity-preserver_README.mdmarkdown6.8 KB
examples/karpathy/data/github/karpathy_char-rnn_README.mdmarkdown12.5 KB
examples/karpathy/data/github/karpathy_llm.c_README.mdmarkdown16.1 KB
examples/karpathy/data/github/karpathy_micrograd_README.mdmarkdown2.4 KB
examples/karpathy/data/github/karpathy_minbpe_README.mdmarkdown9.1 KB
examples/karpathy/data/github/karpathy_nanoGPT_README.mdmarkdown13.5 KB
examples/karpathy/data/github/karpathy_nn-zero-to-hero_README.mdmarkdown6.7 KB
examples/karpathy/data/influences.mdmarkdown4.9 KB
examples/karpathy/data/transcripts/7xTGNNLPyMI.en.txtmarkdown208.7 KB
examples/karpathy/data/transcripts/P6sfmUTpUmc.en.txtmarkdown108.9 KB
examples/karpathy/data/transcripts/PaCmpygFfXo.en.txtmarkdown96.3 KB
examples/karpathy/data/transcripts/SOURCES.mdmarkdown2.7 KB
examples/karpathy/data/transcripts/TCH_1BHY58I.en.txtmarkdown63.3 KB
examples/karpathy/data/transcripts/VMj-3S1tku0.en.txtmarkdown118.3 KB
examples/karpathy/data/transcripts/c3b-JASoPi0.en.txtmarkdown39.5 KB
examples/karpathy/data/transcripts/hM_h0UA7upI.en.txtmarkdown55.4 KB
examples/karpathy/data/transcripts/kCc8FmEb1nY.en.txtmarkdown105.3 KB
examples/karpathy/data/transcripts/l8pRSuU81PU.en.txtmarkdown213.9 KB
examples/karpathy/data/transcripts/q8SA3rM6ckI.en.txtmarkdown102.5 KB
examples/karpathy/data/transcripts/t3YJ5hKiMQ0.en.txtmarkdown52.9 KB
examples/karpathy/data/transcripts/zduSFxRajkE.en.txtmarkdown122.6 KB
examples/karpathy/data/writing/ai-short-peek.htmlother55.7 KB
examples/karpathy/data/writing/biohacking-lite.htmlother41.2 KB
examples/karpathy/data/writing/blockchain.htmlother297.3 KB
examples/karpathy/data/writing/forward-pass.htmlother14.6 KB
examples/karpathy/data/writing/lecun1989.htmlother25.0 KB
examples/karpathy/data/writing/medium.htmlother6.6 KB
examples/karpathy/data/writing/microgpt.htmlother93.8 KB
examples/karpathy/data/writing/phd.htmlother56.7 KB
examples/karpathy/data/writing/recipe.htmlother32.9 KB
examples/karpathy/data/writing/rl.htmlother54.4 KB
examples/karpathy/data/writing/rnn-effectiveness.htmlother79.9 KB
examples/karpathy/data/writing/software-2.0.htmlother235.3 KB
examples/karpathy/data/writing/state-of-cv.htmlother13.3 KB
examples/karpathy/data/x/README.mdmarkdown564 B
examples/karpathy/data/x/curated-tweets.mdmarkdown35.7 KB
examples/karpathy/data/x/posts-original.jsonconfig171.1 KB
examples/karpathy/data/x/profile.jsonconfig470 B
examples/karpathy/data/x/replies.jsonconfig105.5 KB
examples/karpathy/examples/bad-outputs.mdmarkdown6.2 KB
examples/karpathy/examples/good-outputs.mdmarkdown14.8 KB
examples/karpathy/scripts/fetch-data.shscript6.6 KB
examples/karpathy/scripts/weak-model-test.mjsscript15.2 KB
examples/karpathy/tests/prediction-test.mdmarkdown6.1 KB
examples/karpathy/tests/weak-model-results.mdmarkdown12.6 KB
examples/michael-saylor/MEMORY.mdmarkdown351 B
examples/michael-saylor/QUICKSTART.mdmarkdown3.5 KB
examples/michael-saylor/README.mdmarkdown6.5 KB
examples/michael-saylor/SOUL.mdmarkdown13.4 KB
examples/michael-saylor/STYLE.mdmarkdown7.0 KB
examples/michael-saylor/data/influences.mdmarkdown2.5 KB
examples/michael-saylor/examples/bad-outputs.mdmarkdown3.7 KB
examples/michael-saylor/examples/good-outputs.mdmarkdown7.0 KB
examples/michael-saylor/tests/prediction-test.mdmarkdown4.3 KB
examples/steipete/QUICKSTART.mdmarkdown5.1 KB
examples/steipete/README.mdmarkdown6.9 KB
examples/steipete/SOUL.mdmarkdown16.8 KB
examples/steipete/STYLE.mdmarkdown12.1 KB
examples/steipete/data/github/openclaw--openclaw.README.mdmarkdown84.0 KB
examples/steipete/data/github/openclaw--openclaw.VISION.mdmarkdown5.3 KB
examples/steipete/data/github/steipete--Aspects.README.mdmarkdown10.8 KB
examples/steipete/data/github/steipete--CodexBar.README.mdmarkdown8.5 KB
examples/steipete/data/github/steipete--InterposeKit.README.mdmarkdown11.0 KB
examples/steipete/data/github/steipete--Peekaboo.README.mdmarkdown7.4 KB
examples/steipete/data/github/steipete--Tachikoma.README.mdmarkdown3.9 KB
examples/steipete/data/github/steipete--VibeMeter.README.mdmarkdown13.3 KB
examples/steipete/data/github/steipete--agent-rules.README.mdmarkdown146 B
examples/steipete/data/github/steipete--agent-scripts.README.mdmarkdown5.8 KB
examples/steipete/data/github/steipete--birdclaw.README.mdmarkdown15.8 KB
examples/steipete/data/github/steipete--claude-code-mcp.README.mdmarkdown12.7 KB
examples/steipete/data/github/steipete--mcporter.README.mdmarkdown28.4 KB
examples/steipete/data/github/steipete--oracle.README.mdmarkdown28.3 KB
examples/steipete/data/github/steipete--poltergeist.AGENTS.mdmarkdown9.4 KB
examples/steipete/data/github/steipete--poltergeist.README.mdmarkdown13.5 KB
examples/steipete/data/influences.mdmarkdown10.7 KB
examples/steipete/data/transcripts/3-lessons-you-can-use-today.txtmarkdown10.9 KB
examples/steipete/data/transcripts/agentic-workflow-behind-openclaw.txtmarkdown12.6 KB
examples/steipete/data/transcripts/builders-unscripted-ep1.txtmarkdown31.4 KB
examples/steipete/data/transcripts/first-public-appearance-since-launch.txtmarkdown34.4 KB
examples/steipete/data/transcripts/i-ship-code-i-dont-read.txtmarkdown114.7 KB
examples/steipete/data/transcripts/lex-fridman-491.txtmarkdown158.5 KB
examples/steipete/data/transcripts/lex-fridman-clip-ai-agents.txtmarkdown25.9 KB
examples/steipete/data/transcripts/maximize-claude-code-subscription.txtmarkdown25.7 KB
examples/steipete/data/writing/2012-dont-call-willchangevalueforkey.htmlother59.8 KB
examples/steipete/data/writing/2012-hacking-block-support-into-uimenuitem.htmlother183.8 KB
examples/steipete/data/writing/2012-moving-on.htmlother62.2 KB
examples/steipete/data/writing/2012-nsurlcache-uses-a-disk-cache-as-of-ios5.htmlother54.5 KB
examples/steipete/data/writing/2012-pimping-recursivedescription.htmlother55.6 KB
examples/steipete/data/writing/2012-reboot.htmlother50.0 KB
examples/steipete/data/writing/2012-using-subscripting-with-xcode-4_4-and-ios-4_3.htmlother59.5 KB
examples/steipete/data/writing/2014-a-story-about-swizzling-the-right-way-and-touch-forwarding.htmlother112.5 KB
examples/steipete/data/writing/2025-applescript-cli-macos-complete-guide.htmlother110.7 KB
examples/steipete/data/writing/2025-automatic-observation-tracking-uikit-appkit.htmlother104.4 KB
examples/steipete/data/writing/2025-claude-code-anonymous.htmlother54.8 KB
examples/steipete/data/writing/2025-claude-code-is-my-computer.htmlother72.0 KB
examples/steipete/data/writing/2025-code-signing-and-notarization-sparkle-and-tears.htmlother101.6 KB
examples/steipete/data/writing/2025-commanding-your-claude-code-army.htmlother75.2 KB
examples/steipete/data/writing/2025-essential-reading-august-2025.htmlother78.7 KB
examples/steipete/data/writing/2025-essential-reading-july-2025.htmlother73.2 KB
examples/steipete/data/writing/2025-essential-reading.htmlother79.4 KB
examples/steipete/data/writing/2025-finding-my-spark-again.htmlother55.8 KB
examples/steipete/data/writing/2025-introducing-demark-html-to-markdown-in-swift.htmlother63.9 KB
examples/steipete/data/writing/2025-live-coding-session-building-arena.htmlother59.8 KB
examples/steipete/data/writing/2025-llm-codes-transform-developer-docs.htmlother59.9 KB
examples/steipete/data/writing/2025-logging-privacy-shenanigans.htmlother108.4 KB
examples/steipete/data/writing/2025-mcp-best-practices.htmlother69.0 KB
examples/steipete/data/writing/2025-migrating-700-tests-to-swift-testing.htmlother178.7 KB
examples/steipete/data/writing/2025-optimal-ai-development-workflow.htmlother67.8 KB
examples/steipete/data/writing/2025-peekaboo-2-freeing-the-cli-from-its-mcp-shackles.htmlother76.2 KB
examples/steipete/data/writing/2025-peekaboo-mcp-lightning-fast-macos-screenshots-for-ai-agents.htmlother67.1 KB
examples/steipete/data/writing/2025-poltergeist-ghost-keeps-builds-fresh.htmlother81.8 KB
examples/steipete/data/writing/2025-self-hosting-ai-models.htmlother114.9 KB
examples/steipete/data/writing/2025-shipping-at-inference-speed.htmlother70.1 KB
examples/steipete/data/writing/2025-showing-settings-from-macos-menu-bar-items.htmlother110.3 KB
examples/steipete/data/writing/2025-signature-flicker.htmlother69.5 KB
examples/steipete/data/writing/2025-startup-slop.htmlother69.7 KB
examples/steipete/data/writing/2025-stats-store-privacy-first-sparkle-analytics.htmlother58.6 KB
examples/steipete/data/writing/2025-stop-overthinking-ai-subscriptions.htmlother67.4 KB
examples/steipete/data/writing/2025-the-future-of-vibe-coding.htmlother88.9 KB
examples/steipete/data/writing/2025-understanding-codebases-with-ai-gemini-workflow.htmlother62.5 KB
examples/steipete/data/writing/2025-vibe-meter-2-claude-code-usage-calculation.htmlother71.3 KB
examples/steipete/data/writing/2025-vibe-meter-monitor-your-ai-costs.htmlother69.9 KB
examples/steipete/data/writing/2025-vibetunnel-first-anniversary.htmlother59.0 KB
examples/steipete/data/writing/2025-vibetunnel-turn-any-browser-into-your-mac-terminal.htmlother77.7 KB
examples/steipete/data/writing/2025-when-ai-meets-madness-peters-16-hour-days.htmlother70.4 KB
examples/steipete/data/writing/2026-openclaw.htmlother52.1 KB
examples/steipete/data/writing/command-your-claude-code-army-reloaded.htmlother59.4 KB
examples/steipete/data/writing/index.htmlother34.0 KB
examples/steipete/data/writing/just-one-more-prompt.htmlother59.7 KB
examples/steipete/data/writing/just-talk-to-it.htmlother80.8 KB
examples/steipete/data/x/curated-tweets.mdmarkdown42.1 KB
examples/steipete/data/x/openclaw.posts-original.jsonconfig68.4 KB
examples/steipete/data/x/openclaw.profile.jsonconfig478 B
examples/steipete/data/x/openclaw.replies.jsonconfig74.6 KB
examples/steipete/data/x/steipete.posts-original.jsonconfig95.5 KB
examples/steipete/data/x/steipete.profile.jsonconfig485 B
examples/steipete/data/x/steipete.replies.jsonconfig2 B
examples/steipete/examples/bad-outputs.mdmarkdown10.2 KB
examples/steipete/examples/good-outputs.mdmarkdown15.9 KB
examples/steipete/scripts/fetch-data.shexecutablescript11.4 KB
examples/steipete/scripts/pull-x.mjsscript3.9 KB
examples/steipete/scripts/weak-model-test.mjsscript17.7 KB
examples/steipete/tests/prediction-test.mdmarkdown13.2 KB
examples/steipete/tests/weak-model-results.mdmarkdown11.6 KB
examples/vitalik-buterin/MEMORY.mdmarkdown972 B
examples/vitalik-buterin/QUICKSTART.mdmarkdown3.3 KB
examples/vitalik-buterin/README.mdmarkdown6.9 KB

examples/steipete/data/transcripts/builders-unscripted-ep1.txt

View on GitHub
[music]
Peter, welcome to OpenAI.
>> Thanks for having me.
>> We've known each other online now for
many years. Uh, but I'm so happy to
finally get the chance to spend more
time with you in person.
>> Likewise. Beautiful office, by the way.
>> Thank you. Thank you. You had a crazy
couple weeks. We first had the idea of
doing a video video together a month
ago. If we had done it then I would have
had to intro you. I guess you don't even
need an intro anymore. Uh it's not often
that an open source project makes the
WRA journal. Uh so congrats on all the
success. How are you feeling?
>> Slightly some story overload on all
points but also you know when I when I
started out this year like messing with
AI, I wanted to like inspire people and
I feel like this is the final form.
[laughter]
So I'm I'm I'm proud.
>> It's been amazing. You've been in SF for
the past week. Uh, and you attended some
events like the Codex hackathon, but you
also had claw con like an event
dedicated to open claw.
>> Well, it also was created by the
community.
>> Yes. I created this discord channel
about meetups. After people were like,
we need to have a meetup and I'm like
sure. And then I came to Clock and there
were like a thousand people and I was
just blown away by like the creativity,
the
the the
colors, like there are so many things,
so many people excited.
>> I mean, that's how you realize that
you've built something magical. Like the
project did not even exist a few weeks
ago and now you have like thousands of
people embracing it, using it, gathering
to meet you in SF. It's pretty
incredible. Even in Vienna next week, we
have like already 300 people and it's by
far not as much of a tax scene.
>> Wow.
>> As as San Francisco.
>> It's now worldwide.
>> So now it's a phenomenon.
>> Yeah. That's amazing that that this
reaches different continents, different
cultures.
>> Indeed. And and so how has the
conversation been with like the
community here? You spend time with the
community, some of the maintainers also
that you've kind of like brought into
your project. Yeah.
>> Um, how has that been for you this week?
>> It's been quite something.
>> Um, a lot of people love it. A lot of
people have this expectation of this
enterprise ready final project when for
me for the longest time it was like my
my little playground. I mean the whole
year I'm just marveling at what's
possible now. If you're a builder this
is like what a time to be alive.
>> Yeah. What do you think is the most
interesting thing about building like
actually building being a builder at
this moment in time like it's a very
interesting time where like the entire
tool chain is changing
>> the definition of what it is to be a
developer is also changing and anyone
can build anything when I when I started
>> playing with with this new tech I got a
dopamine hit every time back then I
started with cloud code when it got
something right and it was like maybe
30% or 40% chance but I To me, it was
like
like literally mind-blowing because I
had this realization that now I can
definitely build anything and um you
know, usually you're so time constrained
because software is hard and software is
still hard but you're so much faster
now. I agree. Uh so if we take um a few
years back maybe and rewind a little
bit. So you I got first got to know your
work back I think in 2011 or 2012 when
you built PS PDF kit.
>> Yeah.
>> And from the outside it's interesting
because it feels like you've achieved
the dream of every developer. You kind
of saw a problem. You created a great
solution for it. You built a company
around it. You scaled it. You solved it.
But I'm sure that journey is never that
easy.
>> I mean I didn't wake up one day and
thought I'm going to build a PDF
framework. that is like minus 100 on my
interest list. Uh so it kind of kind of
just happened like a weird butterfly
effect from being at the Nokia
development days to having friends that
needed it uh and an American visa taking
too long kind of emerged to me creating
a company.
>> What I find interesting is like after
that company uh you you built you kind
of took a break for a little while.
>> Yeah. So, what brought you back?
>> At the end, I I was just really burned
out. Like, I was running on high steam
for 13 years. Running a company is hard.
Being a founder is hard. It's And I also
didn't really It's my first one. So, I
didn't really had the the knowledge
how to how to mitigate those things. So,
I was just burning a bit too bright and
I needed to decompress. So, I still
follow the tech news. I I saw I think
the very early things of GBT engineer
whatever it was called of chat GBT and I
was kind of cool but
didn't really excite me is you always
have to experience new technology just
by reading from it doesn't really convey
the power
>> so so even that tech didn't really click
and only when I was ready when I felt
the need for like okay I want to build
something again and then I
I didn't want to build it in Apple tech
anymore cuz I I'd done this for such a
long time and
the world kind of moved on a little bit.
You know, you need to things need to be
web first and I I kind of had the hinge
back then, but it's so hard when you go
from when you're such an expert in one
field and you move to another field.
Hard is even the wrong word. It's it's
more like painful.
>> Mhm. Because then you're like you have
all this this broad knowledge of how to
build things
but to really do that without
agentic engineering you need to still
learn a lot to to move your knowledge
over and then I thought like ah let's
let's check out what this AI stuff is.
The moment where it blew my mind was I I
took this project that was halfway done
that where I just yeah burned out before
finishing it basically.
>> It's often the case that like for us
developers we love having new ideas and
starting projects but getting them to
the finish line is the hard part.
>> Oh I see it all the time. It's it's it's
really hard and I even sometimes fail.
But for this one I I wanted to continue
but I also wanted to rewrite it.
>> Mh.
>> So I I took the whole thing. I made one
huge markdown file like I don't know 1.5
megabyte with all the files. I dragged
it into Gemini Studio 2.5 at the time. I
said write me a spec and I had like
those 400 line spec then I dragged it
into cloud code and then I wrote build
and I I did other stuff on main screen.
It was just like running on the side
screen for hours. It was it was things
were quite a bit rougher back then. And
at one point it it told me
I'm 100% production ready, you know,
like the the psychopic uh Opus 3.5 or
whatever it was. And I tried it and it
crashed and then I like hooked up um
Playright.
I think one of the few MCPs I would
actually use and told it to like build
the login stuff and check the work along
the way.
>> Yeah. or something for Twitter and then
1 hour later it it actually worked and
it did show me some stuff. I mean it was
like the worst slop but to me this was
like the moment where it really clicked
cuz like
>> this is like just process-wise I got
like goosebumps for like the
possibilities and that was kind of
moment where I couldn't really sleep
anymore.
>> Yeah. Because by now then then like my
head exploded with all the things that I
I always wanted to build that I just
couldn't before and then I just really
went into the rabbit hole. A lot of
people have seen like Open Claw as an
overnight success for you. But what I
love and I what I find fascinating in
your story is kind of how it's a
culmination of so many projects that
you've been working on for the past like
9 or 10 months, right? Like when you
look at your GitHub profile, there's
like 40 plus projects you've built. Um
half of them I use in the project. Yeah.
>> Yeah. And I think like many of those
Yeah. You you combined them all inside
OpenClaw. Yeah. Can you tell us more
about this journey about like how all of
these ideas and projects made their way
to openclaw?
>> I wish I could say that I had the
unified plan in the beginning but a lot
of it was just exploration and I wanted
things and those things didn't exist and
and I just yield them or let's say I
prompted them into existence. Why? you
know, it's like, okay, let's build it.
>> And then like and it was like step by
step just because I yeah, I wanted my
agent to like do some stuff for me. I I
didn't have that this this this unified
vision yet. I I had a it's funny how it
goes full circle because I I wanted to
have something
that would look at for example my
WhatsApps
and then I built I haven't got the
domain for that. I built the prototype,
but then I thought all the big labs are
going to build it. Like I just wait a
little bit. I focused on other stuff. I
just experimented a lot. I my mission
was kind of like to have fun and inspire
people. And then by November, I I built
a few versions of what I wanted, but
nothing really good. And by November, I
was like, why still why did no lab still
build any of that? You know, like what
the what the heck are they doing? And
then I I built
the first version that became open
class. We at name number five now. Um
still didn't fully click for me. It was
like a this is cool. Like took like one
hour to like build the first prototype
cuz you just you just yield things into
existence now.
And then where it really clicked was
where I was at this uh weekend trip in
Marrakesh and I I found myself using it
way more because it was so convenient,
you know, like there was no there's no
really good internet. So WhatsApp was
WhatsApp just works everywhere. Um and
it was just so convenient. Like I made
pictures for to translate stuff to like
find me in restaurants but also like
look up stuff on my computer. It was
really cool. And I showed it to friends
and I made it like send me text messages
for me and and they wanted it, you know,
and I'm like, "Yeah, you you shouldn't
use it. You don't get that. It's too
dangerous for
>> these are the [clears throat] only signs
of like, you know, uh product market
fit. If your friends already want what
you have, even though you're going to
never design that for them, it was more
like reserved for the technical like
peers that you have."
>> Yeah. And then when it really clicked
for me was I used it so much and at some
point I sent like a voice message
and and I was like wait this shouldn't
work.
>> Oh yeah. Tell me more about that story
because I think we talked about it the
other day. That's fascinating.
>> Yeah. It showed me how good those models
are in problem solving. You know like we
built these things for agent for agendic
engineering but really the skill is like
more abstract. like it's if you want to
be a really good coder, you need to be a
really good problem solver. And that
just maps in any domain really.
>> Yeah.
>> So I I sent this voice match and the
typing indicator appeared and and I'm
like I'm very curious what happens now.
This I didn't build this can't work. And
then the model just replied to me and I
were like how did you do that? You just
like asked the model like how why this
should not work. And the model was like
yeah you sent me a a message but it was
just a file with no file ending. So I
just looked at the file header and found
that it's OPUS the the audio codex. So I
used ffmpeg on my computer to convert it
and then I wanted to to transcribe it
but it didn't have whisper installed.
So I found I looked around and I found
an openi key and I used curl to just
send a file to open ai and got the text
back and here I am. This is incredible.
I mean like that's the power of like
giving tools and the complete access to
a computer to these like agents
>> and now they can just like actually come
up with the solutions themselves even
though you never programmed them at all.
>> It's so funny like I I told the story
and then people are like oh my god it
just used your key like this is madness.
I'm like no like I put it in the
environment exactly for that reason. If
it's this kind of like a script that
should access my open key like my bot
works in the same in the same
environment of course it should access
my open I like I put it there for that
reason. This is not bad. This is exactly
what I wanted.
>> Yeah.
>> Yeah. That's that's where I had my
little moment and then every time I I
showed it to friends or I put them in a
little group chat
like to be to be frank like this this
thing is designed for one-on-one
communication, right? So if you put in a
group chat, pick someone that you trust.
>> Really trust. Yes.
>> Yeah. Because it was not designed for,
oh, let's just put it out there and it
will always do the right thing. You
know, it is your personal assistant.
>> When I set it up, I was also very
intrigued. I was like, this is a weird
setup. I'm like, where is this going to
take me? Uh, but I had also a few like
aha moment where like the more access
you give it, the more tools and skills
you give it, the more impressed you can
become. You also give it like a virtual
skill and then you ask it to build a
website or application even for an event
you want to host. Not only is it
building the app, it's also using your
openi key to like make some AI features
into it. It's deploying it to versel.
There's already a link you can share
with friends. It's a kind of like
mind-blowing step up change versus just
like oh I'm augmenting myself to write
code. throughout November and December I
was like hooked on this and I I mean I
still do some other projects but like
most of my time went there
but on on Twitter
people wouldn't get it and I was like I
got very muted responses and I'm like
every time I showed it to a friend they
wanted it and I'm like no it's not ready
yet you know. So then I was like okay
what's the most insane thing I can do to
like show people how cool this is? So I
created a discord and just put my my bot
in there without any security because
back then I didn't even had sandboxing
built in. It was just very early and I
just worked in the open. I basically
built now open claw with open claw and
debugged it and then asked the model
like oh do you see this tool called no I
see nothing. It's like oh check check
your own source code and and and and
did all a bunch of stuff and people saw
it and then they understood.
>> And when you uh put it in discord like
this what kind of access did you give
it? Did you also give it like all your
tweets for instance? Like what kind of
like knowledge that did it have about
you?
>> Not all my tweets too many. Um but a lot
of my memory stuff. Yeah. So I I
actually was monitoring it very quickly
because yeah prompt injection is
unsolved. But also the latest generation
of models is really good. So I have like
one canary
my soul.mde that kind of defines
what are my values? How do I want the
model to to work to operate to sync?
What's important to me? That's secret.
Yeah.
>> And and people really wanted that and
and I random people came in and tried to
like prompt injected and pasted like
huge blocks of code and the model was
like I'm not reading this. So it was
basically mocking them. But I was still
not very confident, right? So the first
night was like got me a whole lot of
interest. And then I I turned it off,
went to bed, slept like 10 hours, woke
up, there was like 800 messages on on on
Discord. My agent was replying to every
single one. [laughter]
I freaked out. Like I turned it off
again. I read through every single thing
and and at some point I chilled out
because it actually didn't it didn't do
anything malicious. It didn't even
manage to get my solder to be out. I'm
not saying this is I'm sure problem
detection is possible, but it's not as
easy as people think it,
>> right? In the grand scheme of things, it
was actually performing the way you
wanted it to.
>> Yeah. My big mistake was I I disabled
it, but I forgot that I actually had a
launch demon.
>> What What's the the main thing a launch
demon does? If the thing crashes or is
killed, it restarts it.
>> Yeah.
>> Because you want a reliable service and
like Apple built this to be a reliable
service and I I just didn't think about
it. So, I killed it and it restarted
within 5 seconds while I went to sleep.
Now I know better. But now I'm also I
also built in sandboxing. So he was so
proud that it's in his Mac studio. It
calls it the castle and then I put it
into a little little set container.
But you know those models are so
creative. Like the first time I I I made
a I made an LP Docker container and
there's like nothing in it. And I I I
told now Malti like hey can you check
out this website and they were like
there's not even curl. There's nothing
in here. And I'm like, be creative. So,
it built it built um lobster curl just
by like its own tool.
>> It just built its own its own curl by
like using some some TCP socket and like
there was like a C compiler and then it
had like it built like a crappy version
of curl so it could actually access a
website and it worked crazy. So, I'm
like those things are so resourceful.
It's incredible.
>> Yeah. you had some challenges too of
like you know people really like uh
taking a look at the security issues
potentially and expecting you to have uh
everything so robust from day one
although you were just putting an open
source project out there. I always love
when people ask me like, "Oh yeah, can
you put me in with like the CEO or human
resources or some other member of my
team?" And I'm like,
>> "It just me hacking from my cave
basically
>> like but you that's where you see the
dissonance, right? This would have not
been possible
by any human, any one human. Yes, I have
maintainers now and I get PRs, but
essentially I'd built this.
>> Yeah. even a year ago it wouldn't have
been possible. There is no model for
like something like this could be built
by one person. So it's not even they
don't even consider that
>> right. Yeah. Let's actually touch on
that like I think um on your like
productivity because I'm sure like many
developers must be wondering like how is
Peter so productive and I was looking at
your uh GitHub again this morning and I
saw it's like 90,000 contribution on
more than 120 projects just this past
year. But what's also interesting is
like you know the past year started very
white on the GitHub activity graph and
then light green and then into fall
around October November it's very dark
green. Yeah. Um
>> that's where
>> what happened
to Codex [laughter]
>> like uh with every generation they got
better but it's not just that the agents
got better the harnesses got better also
my understanding of how to approach this
and my workflow got better. You know,
there's these people that write software
in the old way and the old way is going
to go away. And they try,
they call it vibe coding. I I think VIP
coding is a slur. And then they try AI,
but they don't understand that it's a
skill.
>> You you you pick up the guitar, you're
not going to be good at the guitar on
the first day.
>> So, they going to have a bad experience
and then they're like, "Oh, no, it's not
going to work."
While if you approach it more like
playful,
it you have to learn. You have to like I
have a gut feeling now for like which
prompt will and how long it'll take. And
if it takes longer, I'll I'll reflect.
Oh, maybe I made a mistake. Maybe my
architecture is wrong, my thinking was
wrong, or maybe something else. Just as
with code, if you if you if you write
code and and you have a feeling of, oh,
this is this naturally goes into my
architecture or this is this is I'm
fighting the the system and it just
takes time.
>> And so if people want to become more
like you, uh what is your codec setup
today? Cuz I think you famously said
like most people over complicate their
setup.
>> I mean I also over complicated my setup.
I I call it the agentic trap from your
first touch point in that new technology
to like becoming really effective. A lot
of people get stuck in there trying to
like super optimize their setup. Doesn't
really make you more productive, but it
feels like you're more productive. So
like I put this one blog post that was
also very controversial, but I just talk
to it. Like you just approach it like a
conversation. Like your model is like
your it's not really pair programming.
It's something different, you know? It's
it's it's a conversation. I basically
tell what I want. I always ask the
model, do you have any questions? for
some reason like there's always
questions about the model by default is
trained to just solve your problem and
then would just make assumptions. The
default assumptions might not always be
the best ones, especially you have to
remember it's trained with a lot of code
and also a lot of older code. Do you
have any question is like a very
important question. People don't realize
that the model usually starts with a
blank slate because they don't learn
like us. Like every new session is like
I know nothing about this codebase and
I'll just search and find the little
things that you ask me to and then try
to solve it. But they don't see usually
the whole picture. If you do this
properly, you have to have the full
picture in your head and you have to
like help the model a little bit to
motivate it to like look here, look
there. And Codex is just much better at
taking a broad look first. I used a very
a very basic approach. I don't even use
work trees. I just have have
check out one to 10. Basically, keeping
it simple helps me to focus more on the
actual problem. So I don't even deal
with branches or work trees. I just
focus on different problems. Ideally,
that gets easier if the project becomes
a little bit larger. Then you can work
on different things that don't fight
each other so much.
>> You used Codex a ton to build open claw.
>> Um, how has Codex changed the way you
work?
>> Well, I tried a lot of tools and what I
like about Codex is that my trust in it
building what I want from all the tools
right now is the highest. And the number
of things that just work is really high.
And I don't think what people realize is
that with GPD 5.2, it was again like a
quantum leap in terms of Yeah, this will
just work. I'm still amazed how well
this already works.
>> That's fantastic. We can just build
things. Uh this is pretty incredible.
>> Yeah, like people need to give this a
try.
>> You famously said also that like you're
shipping code now that you're not even
reading.
>> How is that changed?
>> Most code is boring. like most code just
transforms one shape of data into
another shape of data and eventually the
user sees it or it goes somewhere else.
So like on most most code it writes I'm
I have a pretty good understanding of
what it writes and it's enough to like
that I I see the stream uh and see that
okay like the mental model I have in my
head of what it creates is approximately
what it creates. I was I led a team
before, right? I had a lot of software
engineers under me and that also
requires accepting that they will not
write exactly the same code that I want.
In the end, you you should optimize a
code base so agents can do their best
work,
>> which is not always the same as humans
can do the best work,
>> right?
>> It also means accepting that maybe the
code is not exactly how I would like it,
how I would write it, and I could push
the model towards that. But often times
there's so many ways how you can
structure something that often it
doesn't matter if it becomes a
performance problem. You just focus on
that you make it better.
>> And your point that you just made on
like you kind of like the value of code
and how you approach code is also
changing a lot how you approach open
source right because like I was looking
at like open cloud you now have like
2,000 PRs open right now.
>> Yeah.
before before we had like AI, you know,
you would have to read all of these PRs
because there was value in the code
being created. But sometimes now you
refer to that as like not a pull request
but a prompt request almost because like
more the idea or the intention behind
the PR that's more important than the
code itself. PR is actually sometimes
take me longer than if I would approach
it myself
>> because
my trust in the model not being
malicious is much higher than an
external contributor that I never heard
of before and where we had no discussion
before. So I have to scrutinize that a
little bit more. But then if I see a PR
my like I I start the review and my
first question to the model is do you
understand the intent of the PR? Because
I don't really care about the code. what
they care about what is that person
actually trying to solve. You know, it's
more like an issue with like a here's my
attempt of a solution, but first of all,
many many people still don't really know
how to deal the agent.
>> And then often it's just a very
localized solution because they they
don't have the full system in their
head. The hard part is how does this
little new feature fit into my whole my
my bigger system or how does this little
fix okay, this little fix fix is a tiny
thing, but is it even the right fix? Is
it is it it's probably often this like a
more systemic or an architecture issue.
So the model is actually really good if
you you just have a conversation and
then and when I say okay now build this
then it will start to build this. So I
asked the model what's the what's the
intent? Is that the most optimal
solution?
Sometimes it says yes. Most often it
would say no. Then I would start to
explore what would be the what would be
the best fix. Is this like an
architectural problem? Is this something
that is let's say it's it's a problem in
message handling. Does this only affect
WhatsApp but maybe also I don't know
signal.
>> Yeah.
>> So so should we should we solve this in
a more general way? Is this a new
feature? Do we even want a new feature?
Um, so sometimes those discussion go on
for like 10 15 minutes. I I use like
voice because it's literally like you
you talk to a very smart coworker.
>> Easier to give tokens of a voice than
like typing.
>> Yeah. When when I'm happy then I have
like one slash command like land PR that
explains the whole procedure to like
actually create a branch, do all the
changes, get the PR merged. I I wanted
to create a community. I I try to still
credit the person that created it even
though the whole thing takes me longer
than if I would just write it all
myself,
>> right?
>> Um but I appreciate that people
want to be part of this. What's your
vision for Open Claw now going forward
with all of these contributors that are
kind of like surrounding the project and
and do you see yourself as kind of a
trailblazer for this idea of like a what
a personal AI agent should look like so
that you know a billion people can one
day use something like that?
>> Yeah. I want to find the balance between
this is something that my mom can
install but also this is something
that's fun and hackable which is
difficult. You know, most open source
projects you like download the package.
But for the longest time, my default
install was get clone, build, run, and
then you literally have the source code
on disk and the agent sits in the source
code and is aware of the source code.
And if you don't like anything, you you
literally prompt the agent and it would
change itself like
>> like actual selfmodifying software. So,
a lot of people that never sent me a PR
sent me a PR. That's also why it's more
like a prompt request because they just
lack the understanding of how to build
software that lasts and at the same time
the whole world is is or the whole
security industry is like putting their
eyes on it which has been interesting
and also a little bit frustrating
because it it misses some nuance right
for example the the web server that I
have that's meant for you initially I
built it for debugging and then I made
pretty but it's only meant to be
accessible in your network in your
trusted network but because it should
also be like a hacker's hacker's
paradise there is an option how you can
like change that because like some
people have weird setups maybe they use
like angro maybe use reverse proxies so
there is a reason why I didn't want to
like
limit that but now you have like people
that put it on the open internet even
though I'm screaming in the security
document please don't do that that's not
that's not what it's intended for and
then I have the security people that
point out that oh yeah it doesn't have
login limitations you know like it
doesn't have all these things that you
need when it's on the public internet
like yeah I didn't build it for that
with that intention but because it is
configurable it totally counts as a as a
CVS10
uh so yeah I
I struggled a little bit with that uh
but now I I brought on actually a
security expert that's the main focus I
I realized is I cannot stop people from
using it in ways it was not intended. So
my my big focus now is to like support
all these use cases and help people not
shoot themselves in the foot.
>> Right. And that's the beauty of open
source. People can embrace it and like
come up with ideas that you didn't even
think about.
>> Yeah. That's the beauty and the madness
of it.
>> Yeah. Zooming out maybe for a second
beyond open claw. Like I've talked to a
bunch of developers this week um who
knew you were coming to the Codex
hackathon and told me like how many like
how come does like Peter come up with
all these great ideas like how does
Peter come up with so much creativity? I
don't know if you have any answer to
that or if it's more like your own
curiosity that you're following. It's
more like this realization that
stuff's easy now. So even if I find an
open source thing that maybe solves my
problem to 70%. I just built it myself
which would have been absolutely
impossible a year ago even and now it's
just like I just prompt it then it's on
a second screen and and code works.
We're both from Europe. When I travel
outside of SF and I go back to Europe,
I'm sure you feel the same way like many
developers and engineers have not
embraced codecs and agentic tools just
yet. What would be your advice for them
like as they get started and like should
rethink the way they work and their
workflow? My first advice always is like
approach it in a playful way. Build
something that you always wanted to
build. Um if you are at least a little
bit of a builder, there has to be
something on the back of your mind that
you want to build. Like just just play.
You just have to approach this this this
thing playful cuz like I think the
Nvidia Nvidia CEO said that in the near
term you're not going to replace by AI.
You're going to be replaced by someone
who uses AI.
>> Yeah. Better than you. Yeah. But if your
identity is I want to I want to create
things. I want to solve problems. If
you're high agency, if you're smart, you
will be in more demand than ever.
>> Yeah. Well, it's such a great time for
builders who embrace these tools, shape
their curiosity, and really can bring
any ideas to life the exact same way
you've done with all of these great
projects and and OpenClaw.
>> I think in a year this is going to this
is going to explode.
>> Yeah. 2026 will be an interesting one. I
think it's a fantastic way to end. Thank
you so so much, Peter, for your time.
Uh, wonderful to spend time with you. We
all at OpenAI love your work. We love
supporting builders like yourself and
honestly, you're a true inspiration for
the broader developer community. Thanks
again and we can't wait to see what
you're going to work on next.
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 194kUpdated Claude CodeCodexVerifiedInstall