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/maximize-claude-code-subscription.txt

View on GitHub
What's up engineers? Andy Debb Dan here.
This question hits my inbox every single
week. After every video, engineers ask,
"Can I use my claw subscription instead
of an API key? I already pay for it, so
can I just use it?" Short answer, yes.
And I'll show you exactly how. The long
answer here is there's one rule to
follow and one confusing thread over
time of engineers that work at Enthropic
going back and forth on what we can and
cannot do. and it's all been exposed by
Open Claw. We're going to cover this,
but we're also going to make a couple
things really clear. Breaking the one
rule I'm going to share is how you get
instabbed. This might sound extreme, but
getting banned from Claude could change
the trajectory of your engineering
career. In the Age of Agents, losing
access to state-of-the-art models and
tools will set you back. On the other
hand, we don't want to waste money on
expensive Opus 4.7 tokens when you've
already paid for a subscription that you
could be using. So before you do
anything clever with your Clawude Pro or
Max subscription, you'll need to know
the exact lines Enthropic drew up in
their latest clarification. The exact
lines you do not want to cross. To make
things absolutely clear here, we'll
break down the claw subscription into
three tiers of usage. safe,
controversial, and bannable. Here's how
you can maximize your Claw subscription
without getting banned.
So, here's the whole video in 20 seconds
so you know exactly what you're getting.
First, we're going to cover the one most
important rule to follow. Then, we'll
break down the three tiers of usage,
what you can do, what's controversial,
and what gets you banned. Here's the one
rule that can get you through 90% of all
terms of all usage. Memorize this one
line here. Your Pro or Mac subscription
is for your individual use. The moment
your code routes someone else's request
through your subscription, stop using
the subscription oath token and switch
to an API key. The second another human
becomes the intended user, you've
crossed from individual use into product
territory. And that's the big
distinguishing factor here. If you just
search product on their usage policy,
developers building products or services
that interact with cloud's capabilities,
including the SDK, should use API
authentication. So we're saying use an
API key. The second what you're building
is not just for you. So this is the
golden rule. It's for your individual
use. The moment someone else is using
your subscription, you have violated
terms of service. And so the test in one
sentence is, am I the only human whose
work these agents are running? If so,
ooth off key and your subscription is
fine. If not, use your API key. That's
the whole decision here. Let me make
this super clear here. This is my direct
understanding of all anthropics
published documentation as of April
2026. If you work at Anthropic and I got
anything wrong, email me. I'll amend
this video immediately. The goal here is
fair use and alignment for every single
engineer watching, nothing else. So what
is safe? What can you do with your claw
subscription? Let's start with safe use.
Personal scripts on your own laptop.
Cron files, pipelines, agentic
workflows. That's all safe use. You are
the only user there. The cloud agent SDK
running your own agents doing research.
This is all valid fair use. CI running
your own repo with the cloud code o
token set. Again, fair use as long as
you're the only one on this repo. Okay,
that's important. cloud code on your
work laptop for engineering. You author.
So, you're writing code, you're building
products, you're doing whatever that you
are authoring. This is still safe use.
You're all good to go here. Building a
product is fine. As you'll see, using
your subscription inside of the product,
that's not fine. Cloud code web, cloud
code desktop, all the cloud code related
functionality, features, tools, apps,
obviously that's okay. Enthropic wants
you to use their tools. All right, so
that's super clear. The big idea here is
one human, one subscription, one
beneficiary. Now, if you have a team or
enterprise subscription, it's more about
your team usage. The rules are a little
different there. But for individuals,
it's one person, one subscription, one
beneficiary. If you remember nothing
else, remember this. This is all safe
use. Now, things get a little murky.
Let's look at the controversial usage of
your claw subscription. Okay, this is
where it isn't clear and where you start
blurring the lines a little bit. agency
or contractor work using your personal
token. And so here we're talking about
workflows that use your OOTH token
through your subscription that's exposed
to your agency or your contractor work.
Okay, I'm not talking about writing the
code. I'm talking about a service that's
ongoing where you are not the only
person benefiting from it. So this is
where things get a little more
controversial. Slack bots daily reports
used by multiple humans. Okay, so a
little more controversial here, right?
Because the output is not just for you.
we've broken that rule about who the
intended output is for. This is where
things get a little weird because when
you're writing code, when you're
building a product, that's all fair use.
But the moment the output and the usage
is shared, it's not so clear. Open
source CLIs only safe if everyone brings
their own token. If you're embedding
your O token into a website or some CLI
tool and you're providing usage through
that interface using your token that
multiple humans are benefiting from that
is controversial use. I would steer away
from that type of usage. Internal team
tools running on one developers pro or
max token. This is very risky. We're
getting into that zone where it's
clearly multiple people benefiting
inside the process, right? Not just the
output of the thing, the process of
building the thing. Internal team tools.
Again, this is a little different if
you're running an enterprise or team
subscription. That's safe for that.
You're going to want to reach out to
Anthropic directly. For individuals,
this is crossing the line. Blurry, but
we're crossing the line. This is
something you could air quotes get away
with given your token usage is low, but
I would not recommend that. Here we just
switched to the API token. And then the
most controversial of all, third-party
agent harnesses, open claw style. This
is highly contested. Let me just wrap up
here and then we're going to refocus
back on this because I thought diving
into this I would be able to just jump
in, find the clear answer, share it with
you guys here today. But I spent a lot
of time and frankly quite annoyed with
how much time I've had to spend tracing
through Enthropics, flipping back and
forth and back and forth on thirdparty
agent harness usage. So we got to talk
about this a little more. But overall
for controversial uses, right,
everything I've mentioned here, I just
recommend you grab the API key here and
stop guessing. Okay, this is the winning
move here. Don't risk losing access to
some of the most powerful, most
intelligent compute and tools.
Regardless of what you think about
Anthropic, they are your interface to
some of the best intelligence on the
planet. Okay? So, don't burn that
because you're trying to save a few
bucks. It's not worth it. I will also
say here, as you'll see in a moment,
Enthropic needs to be clearer about
these use cases. I found myself just as
confused as many other engineers, which
is a serious problem. So, let's go ahead
and just understand this. What is
anthropic's position on using the OOTH
for personal work inside of third-party
agent harnesses like OpenClaw? Let me
just kind of share how confusing this
has been walking through this and try to
provide some clarity as much as I can
here. So, let's start from the beginning
or one beginning. So February 18th, Tar,
one of the anthropic team members says
here that we want to encourage local
development and experimentation with the
agent SDK. If you're building a business
on top of the agent SDK, uh you should
use an API key instead of the OOTH key,
right? Instead of your subscription.
We're going to make this clear in our
docs. Okay, fine. So this seems like a
really clear start. Next, Boris
Churnney, the creator of Cloud Code,
says tomorrow 12. So this is April 3rd.
Okay, so we're moving a little bit into
the future. All subscriptions will no
longer cover thirdparty usage like
OpenClaw. Okay, fine. Third party usage
is not allowed at all. Okay, so this
means that we cannot use agent
harnesses. Third party agent harnesses
like OpenClaw. To Boris's credit
completely, we've been working hard to
meet the increase in demand for Claude.
Our subscriptions weren't built for the
usage patterns of these thirdparty
tools. Uh capacity is a resource we
manage thoughtfully and we are
prioritizing our customers using our
products. Makes sense. There's nothing
confusing about that. If you're compute
constrained, which it sounds like they
are, you want to have all the demand
coming through your products. Okay, so
nothing wrong about that. This is where
things get a little annoying and weird.
April 6th, we have Peter Steinberg, the
creator of Open Claw, says Anthropic is
now blocking first party harness use as
well. Laura says, "This is not
intentional. Likely an overactive abuse
classifier looking and working on
clarifying the policy going forward."
Okay, if we move forward here on the
10th, Peter got banned from OpenClaw
usage and then he got his account
reinstated. We're kind of flipping back
and forth here. I'm not sure if I can
use OpenClaw, if I can use Open Code, if
I can use the PI coding agent. It's not
clear what I can use here. And then on
the OpenCloud documentation, Enthropic
staff told us OpenClaw style CLI usage
is allowed again, but for long lived
gateway hosts, anthropic API keys are
still the cleanest and most predictable
production pathway. We're flipping back
and forth here all over the place. There
are many, many, many more tweets like
this that say you can use it, say you
can't use it. You know, you shouldn't,
but you can. And we're unbanning Peter.
This has just been a nightmare to track
down. And you know, like I said, there's
like a million more tweets like this. I
think it's obnoxious that I have to
spend my time as an engineer going on
Twitter to try to find the source of
truth for what should be in the usage
policy documentation. And if you come
back to this, as we should be, because
this is where the real legal bounds
actually lie in the usage policy here.
Here's what Anthropic is saying, and
this is what I like to go by when I'm
building. This is what I do. I look at
the documentation. I don't look at
tweets. I don't look at what people are
saying. I look at what the company says,
which can be quite challenging because
very clearly they're flip-flopping here
a decent amount on how you can actually
use their tools. Okay. But I think the
definitive guide is here. Again, as
mentioned, developers building products
or services that interact with cloud's
capabilities, including the SDK, should
use an API key. So, a couple different
metrics you and I can use. If you're
building a product or service and you're
servicing more than just yourself, use
an API key. As mentioned, that is the
golden rule here. If you're building a
product or a service or other users
requests are being routed through your O
token, do not do that. Okay, that is
violating the usage policy as stated
here. Of course, full disclaimer, read
and look over the documentation yourself
and try to track all the tweets as well.
If you're working at Anthropic, please
make this clear. Add explicit examples
of what we can and cannot do and please
address third-party agent harnesses.
This is how I interpret the
documentation. Again, you know, do your
own research here. Given the current
state of things, I'm currently not using
my OOTH token, which we're going to
break down how to gain access to and how
to use in just a moment here. I'm not
using that for any third party tools. My
favorite agent harness right now is the
PI coding agent. This is my go-to for
fully controlling the agent harness. I'm
a huge Cloud Code fan. I've been using
it since the beginning, but it has
become more and more bloated over time.
I'll link a big hitting video where we
talked about the PI coding agent versus
cloud code in the description if you're
interested in that. The big idea here is
you want an open- source cloud code
competitor that you can fully control
down to the font color. And this is
going to be really really important as
we move further into multi- aent
orchestration and as we scale our
compute to scale our impact which is
typically the focus of this channel. So,
if you're new, like and subscribe to
stay on the edge of aic engineering.
Normally, we don't cover what I like to
call configuration level issues like
this. But it's come up over and over and
over. And as I started looking into
this, it's really not that clear. So, I
wanted to address it directly here on
the channel. What is gray here is open
claw style third party agent harnesses.
This is not clear at all cuz I'm telling
you right now I think the bigger issue
here is and we'll talk about this at the
end. It's becoming more and more clear
that Anthropic is compute constrained
and they're trying to first service
their highest tier most important
subscribers which makes sense. That's
exactly what you would do. You want to
defend your high income revenue users,
right? It makes complete sense. You
cannot blame them for that. But what we
can blame them for is not being clear on
exactly how we can use the subscription.
Obviously, as a longtime Claude
subscriber, I have the Mac subscription,
and you know, I've been paying them 200
every single month for over a year now.
Obviously, I want permissive use. I want
to be able to use this all the time,
everywhere, right? But that's not the
reality of running a business. And so, I
understand that, but what I wish they
would do is just make it clearer. This
is how I'm interpreting this. I'm
playing it safe. I'm definitely on more
on the conservative side here on trying
to like hack out tokens and try to use
the subscription for every little thing
because I realize the importance of the
tooling and specifically the models
coming out of Enthropic. That's the
controversial use and this is the
biggest one. Make sure you follow the
channel, drop a like, let the algorithm
know you're interested. I'll be updating
any changes here on the channel around
the controversial use of the clawed
subscription. Now, let's move on to the
final tier. If you do these things, you
will be banned and it's really only a
matter of time. someone at Anthropic
will find you and more likely their
agents and their abuse detection systems
will find you and ban you. If you're
shipping a product that runs on your Pro
Max O token, you can get banned and it's
very easy to detect this. How many
tokens are coming through your service?
What's the classification of the prompts
that are running and are you running in
a third party agent harness with a bunch
of random requests that a single
individual user could not do? It's kind
of easy to detect this guys. So like
don't be an idiot. This is banable use.
Multi-tenant SAS application logging in
to claude on your behalf obviously
banable pulling your one subscription
across a team with no team or enterprise
seats banable reselling claude banable
instant ban okay extracting tokens
pier.json JSON or your keychain and
using them. Banable, right? This all
clearly abuses of your subscription.
Don't do these things. Don't misuse
software. Don't be malicious. It's in
your benefit to not get banned. Right?
My oneliner here is don't trade Frontier
AI access and Frontier AI tooling for a
few hundred bucks. You're trying to save
money. You're trying to build a
business. You're trying to do your
thing. I get it. But you don't want to
lose access to premium compute, to
premium intelligence. This will get you
banned full stop. So, now we have a
better idea of how to use a
subscription. Let's break down exactly
how to use your OOTH token and not your
API key because there's one big gotcha
here that if you don't catch this, you
will be billing your API key and not
your OOTH token.
So, I have this codebase here that's
going to be available to you. So, in
this codebase, I have a few very, very
simple but very informative examples of
how to set up your OF key in your
systems. And again, I'm assuming you're
following the one rule so that you are
not breaking your subscription usage
terms, okay? Your own use and you're not
routing someone else's requests. By the
way, for gray controversial usage, I
didn't mention this. I want to clearly
state that if you're unsure, you can
just directly contact support. They do
mention this or contact sales. Actually,
for questions about permuted
authentication use, please contact
sales. I am going to ask them directly
after I film this video if I can use a
third-party agent harness like the pi
coding agent. I'm just going to directly
ask them and see what they say because
this is my favorite agent harness. I'm
absolutely not an open claw user. I do
not recommend using that tool. Uh that's
a topic for another time, but I'm going
to go ahead and directly ask them. This
is something that you can do too, right?
If you're not sure about that gray area,
it's always just better to ask. So, let
me show you exactly how this works. So,
I'm using just this is a simple command
line runner and it runs out of a just
file. So in this file, you can see we
have a bunch of simple commands. They're
mostly just routing to Astral UV, the
best way to run Python. They're just
kicking off single file Python scripts
with concise examples on how to use your
OF key versus your API key. If I type J,
which is an alias for just it's going to
just show me all the commands I have
available. We have the CLI and the agent
SDK. You can run your OOTH and your API
token directly through the CLI. So, you
know, we can do claw-p ping, right? And
this is going to run directly to the CLI
and you can set up through the agent SDK
that's available. Let's run this J API
CLI. So here we've used the API key. The
way that you can know that you're using
your API key is this field here API key
source. And you can see here rate limit
events. The API key has no rate limit
event. This is going to differ from what
we'll look at in a second which is your
OOTH usage. Right? Oth usage bills
against your subscription which has rate
limits just by coming into this codebase
running the install command which I set
up in every codebase now. So if you go
into whatever cloud code instance you
have or any agentic coding tool really
you don't need cloud code you can do
/install and I have this prompt here
that's going to automatically help you
set up this codebase very very quickly.
After you get that all set up you can
just type jpi cli. It'll run everything
you need to see and understand to know
for a fact using your API key or your
ooth token. So all the code is here in
this API key CLI file. But the important
piece is here. So if we go into the
output ind JSON file, this is just the
raw output stream of that execution in
the log file. You're going to know for a
fact that you're running your API key
through this field. If you just search
API key here, this is the field that
lets you know for a fact you're running
your API key. So very important, you
know, we got build. If we look for
total, we can see our total costs. 2
cents for a ping prompt. Pretty crazy,
right? That's because of that massive
system prompt. But if we back up here,
type J again, and we do J O CLI, this is
going to run our subscription. Again,
this is all going to be in this codebase
for you, link in the description. It's
going to be completely free, so you can
understand exactly how to set up your
Cloud subscription using your OOTH
token. We have our events indon file.
And if we go into this, this is just the
raw streamed output from that execution.
Uh you can know for a fact that you're
running your subscription like this. So
if we break this into new lines and we
search API key source, it will say none
here. And it'll also give you this
field. You also see this rate limit
type. On line three, you can see here we
have five hour rate limit type. This is
how you know for a fact you're using
your oath token through your
subscription and not the API key. Now I
mentioned that there's a gotcha. There's
a big thing you need to make sure you're
doing here. If we go into 01 off CLI
line 103, you need to make sure that you
unset your anthropic API key. You need
to make 100% absolute sure that you
unset this thing. Otherwise, it's going
to override your OOTH token. So, you can
see here right in the environment
variable, I'm popping this out.pop. Make
sure you can still read code. If you
can't read this code, probably should
not be using this code. But if you hit
enter here, you can see subprocess.run.
This is the environment we pass into the
CLI. There is an agent SDK version here
as well if you want to use that. But
it's going to make sure that variable
specifically the API key is not set up.
Let me backtrack a little bit. To obtain
your off token, all you need to do is
type claude setup token and it's going
to open up a browser for you and you're
going to be able to get your off token
like this. This is going to build right
against your subscription usage. And of
course, just take that and you're going
to want to set that directly in your
environment variable. I set up both here
just as examples. Cloud code ooth token.
This is the variable name. And then you
have your anthropic API key. So that's
it. It's that simple. This codebase is
going to be available to you. Link in
the description. It contains some more
information and research on the three
tiers of usage. You really want to stay
in this green safe area. Clean setup
guide here. You can go the agentic
installation or manual installation if
you want to make sure you know what
everything is doing, which I always
recommend. Looking through this and
preparing for this video really really
made me think more about what's going on
at Anthropic. Like why is there so much
back and forth and lack of clarity
around how we can use a subscription
that we pay for? And I I think it comes
down to like three things. Anthropic
continues to experience exponential
growth and with that comes massive
growing pains. It's pretty clear and
there's some research and and sources
that state this. Uh at this point they
are compute constraint and they don't
want subscriptions being abused and
overused. That's completely
understandable. But what's not
understandable is the miscommunication
or the unclear or maybe intentional
confusing communication across a bunch
of Twitter accounts. like it's just a
nightmare to track this down. This is
not where I want to be going to know the
tools I can and cannot engineer with. I
really really hope they clarify things.
Again, if you are at Enthropic and
you're watching this, I would love and I
speak for other engineers as well here
would love some more just clear
communication. Give us concrete
examples. Really break it down for us.
What can we do and what can we not do?
The controversial usage just shouldn't
exist, right? You shouldn't have to
guess at what you can and cannot do. I
would love clarity on these items
directly. I think there's a decent
inspiration here for Anthropic to get
this figured out because right behind
them on their tails is Sam Alman at
OpenAI just kind of like giving free
reign to everyone to use codecs and the
GPT models. This news just dropped
today. Apparently, Antropic is testing
preventing any clawed code usage on the
pro subscription. If you scroll down on
this post, Sam Alman comments, "Okay,
boomer."
which just crazy for a CEO at a company
to be saying saying something like this
on someone's post. Um Amal of course
works at Anthropic as well. But um yeah,
they're just kind of getting roasted
here on reducing usage. But I you know,
again, I get it from a business
perspective. You have to depend your
compute resource. It's going to be the
most important resource anyone has very
soon. Probably already is, frankly. So
they're testing blocking cloud code
usage on the pro subscription. Pretty
crazy. Great note here from Simon
Willis, one of my favorite engineers.
This doesn't clarify anything. Please
don't play games like that with AB
tests. Kind of agree with him. I don't
know why this is an AB test. They're
testing sentiment, I guess, around
losing capability. I think it's pretty
clear how that's going to go, but shout
out Sam Alman. This guy's pretty punchy.
I know that the answers here haven't
been 100% clear, 100% perfect. That's
just because there is some gray area
here with the anthropic claw
subscription. They haven't clarified it.
I would love clarification. We'll wait
on that. Make sure you like, subscribe,
comment, all that good stuff, so you let
the algorithm know you're interested.
I'm going to be tracking this
development on the channel to make it
super super clear for engineers who are
just trying to build who's just trying
to create products who are just trying
to master Gentic engineering in the age
of agents. If you're interested in that
and if you're interested in pushing what
you can do with a Gentic technology,
follow that's our bread and butter here
on the channel. I'm not super amped,
frankly, to be covering stuff like this,
but it is important. It is useful. It is
valuable. So, I wanted to bring it here
to you on the channel. I'm playing this
more on the conservative side. I use my
API token right now inside of the PI
agent harness when I want to touch the
anthropic models. I would lean further
away from these controversial uses and
more toward the, you know, safe, very
clear uses. One human, one subscription,
one beneficiary. Don't play with your
access to the most important technology
of our lifetime. With that being said,
your Max Plan is already paid for. So,
let's use it right. You know where to
find me every single Monday. Stay
focused and keep building.
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