Antoine

🚀 Shipping features while escaping 9 to 5

🤖 Building hard

🛠️ More startups to come

[Antoine @onemorelaunch]

Meta: “Introducing Muse, your personal AI agent”
Grok: “Cute. We already have one that doesn’t report everything back to Menlo Park.”
Game on.

[Alexandr Wang @alexandr_wang]

1/ today we're rolling out Muse, our new personal ai assistant. Muse is always-on, wicked fast, can use a browser, connect to your apps, and is designed to be secure.

try it now: https://t.co/n7swQh9v6C.

[Antoine @onemorelaunch]

@falistos @infisical Ça marche super bien, tu renseignes que les variables dans infisical et ça push automatiquement dans ton backend! Tu ne maintiens plus qu’une seule interface

[Antoine @onemorelaunch]

@mateodr_ Meituan pour te faire livrer ta bouffe à n’importe quelle heure!

[Antoine @onemorelaunch]

@tivollem @elonmusk très probable, mais il va se confronter à des régulations et des freins énormes 😤

rien que l’intégration de X Money dans X peut faire peur à pas mal de monde…

[Antoine @onemorelaunch]

the vision is much bigger than i first thought.

i’m building a leaderboard for agentic domains.

every scanned domain gets a score, ranked across categories.

the goal: climb your category by following the recommendations we give you in the app. 🧗

[Antoine @onemorelaunch]

Stop chasing the shiny tool available on the market

Mine is:
⛃ Postgres (Neon)
💽 Next.js
>_ Claude
🌐 Vercel
☁ Cloudflare

What's yours?

[Antoine @onemorelaunch]

I don’t believe, with a few companies there are always issues! Always!

When you book a train ticket with @SNCFVoyageurs I’m pretty sure I gonna have trouble!

When choosing @Chronopost, I’m pretty sure my parcel gonna be lost… or stolen.

What else? Worst services ever!

[Antoine @onemorelaunch]

You’re not paying for tokens. You’re paying for inference time. 🤑

@MistralAI just bet the company on the same insight.

Their latest move isn’t another model. The model race is noise.

The real game is how efficiently those tokens actually run. 🎯

[Kai @hqmank]

Anthropic published "Maximizing the value of your Claude Code sessions". Worth reading in full.

They share 6 tips: /clear between tasks, set model + effort early, @-mention files, quiet flags on commands, /context to
audit startup payload, /compact before you walk away.

It's about Claude Code, but the tips work in Codex or any coding agent.

https://t.co/gz3D5rPXNA

[Antoine @onemorelaunch]

@ericzakariasson Does this loop work with any agent harness on the market, or is it pretty tightly coupled to Cursor cloud agents + PostHog MCP right now? Would love to know how portable the setup is

[Antoine @onemorelaunch]

@elie2222 Well done! @bot should hire you

[Antoine @onemorelaunch]

@uaghazadae what prompt did you feed Codex with? I really like the design he came with! 👏

[Antoine @onemorelaunch]

Most indie hackers fail because they try to serve everyone

Broad TAM feels safe. Laser-focused niche feels risky.

But the ones who actually get traction usually pick a tiny slice and own it completely.

Still stuck on this myself though.

What’s actually worked for you?

[Antoine @onemorelaunch]

@luke_clancy1 Hey! I’m in the analytics space for 5+ years in the corp. world. I want to build something in this space. Your idea might seem very interesting! Can you DM me? (I can’t!)

[Antoine @onemorelaunch]

.@X open algorithm weights “copy link” at 20 ⚡️

A like is only 0.5

One person saving your post to send later beats 40 likes 🥶

Conclusion: write things people actually want to share outside the app.

[X Open Source @XOpenSource]

https://t.co/soxNY8YdP7

[Antoine @onemorelaunch]

I get confirmation when a giant like @Cloudflare named one of its beta product the same way I've had named my product in my repo...🫨

[Antoine @onemorelaunch]

@DanielSmidstrup Plumber! They gonna be augmented by AI but not replaced

[Antoine @onemorelaunch]

Local only, nothing in production. But it finally pushed the proper separation of test and dev databases to the top of the priority list.

Now building databases separation 😮‍💨

[Antoine @onemorelaunch]

Yesterday while working on the annual plan:

After fixing a bug we ran the usual tests to verify.

Because test and dev still pointed to the same local database (a known risk we hadn’t fixed yet), the tests wiped the real dev database including the two Stripe test subscriptions

[Antoine @onemorelaunch]

The session stopped as soon as it noticed something was wrong, checked what was lost, confirmed Stripe itself was untouched, and flagged it instead of continuing on a broken base.

[Antoine @onemorelaunch]

@OpenRouter When I'm asking the session confirmation:
All traffic for this Claude Code session goes through OpenRouter (so I can't see your console or real-time billing).
The current conversation is indeed taking place on deepseek/deepseek-v4-flash-0731
Need explaination @OpenRouter

[Antoine @onemorelaunch]

Explain me?
My local config is
export OPENROUTER_API_KEY="XX" # ligne 8
openrouter() {
ANTHROPIC_BASE_URL="https://t.co/bDrKJ4isuz"
ANTHROPIC_AUTH_TOKEN="$OPENROUTER_API_KEY"
ANTHROPIC_API_KEY=""
CLAUDE_CODE_DISABLE_UNKNOWN_MODEL_WINDOW_ENFORCEMENT=1 \claude "$@"
}