Austin

go with the flow

[Austin @austinrath_]

As a Texas alum and Indian American this one hits home.

Clearly this politician’s viewpoint has resonated with a lot of individuals. More than I would like to admit.

To my fellow Indian Americans, understand that this comes from a place of envy. The people who support this are jealous of the success that we’ve had that they have not.

2nd gen Indian Americans (which everyone in this picture is) are genuinely some of the smartest, most capable individuals in the next generation of Americans. We’re going to be one of the most successful diasporas in the history of this country. Having haters means we’re doing something right.

God bless and hook ‘em 🤘

[Bo French @bofrench]

I heard UT graduation this year looked like this. I didn’t believe it. The problem is now obviously far worse than anyone imagined.

[Kevin He @kevr0n]

@NYCMayor will somebody please get these beggars out of here?

Day 6 of tweeting into the void until I can see the pavement on Canal St again

[Garuth Acharya @GaruthAcharya]

Our adversaries made a simple bet: America would stop believing in itself.

The dream was dead. The factories were gone. Our best days were behind us.

They bet wrong.

Build companies. Build factories. Build families. Build America.

Our best days are ahead. We must build. 🇺🇸

[Garuth Acharya @GaruthAcharya]

We make critical structures for the western world. The AI we're building runs on data you can only get by being on the factory floor and not in ivory towers.

I'm hiring software and manufacturing engineers. Team is ex @SpaceX, @PalantirTech and @DeptofWar.

If this fires you up DM. Time to build.

[Austin @austinrath_]

@GaruthAcharya This guy is fired up !

[Austin @austinrath_]

@n0w00j Too much money in the venture ecosystem.

[Austin @austinrath_]

@deedydas I wouldn’t call it disgusting. Just humans acting rationally.

FAANG companies big draw is good pay and WLB. The 2 year cycle time is a feature not a bug.

[Austin @austinrath_]

Waddle showing everyone how early the harness engineering space is.

For highly complex, specific workflows you’re much better off engineering a harness on top of frontier models instead of RL or fine tuning your own.

Making a good harness needs around a few hundred good examples. RL needs tens of thousands.

A high quality harness solves the sample efficiency problem.

[Hanming Ye @DozenDucc]

Introducing Waddle Labs: Claude Code for robots.

Connect our API to your robot and enter a prompt, then our agents write code to achieve the task in 20 minutes.

@yiding_song @theWaddleLabs

[Austin @austinrath_]

Bearish

The SF economy is propped up by venture dollars being passed around as revenue.

Real engineering happens down in the peninsula, but its share of venture dollars decreased. That tells me everything I need to know.

[Guillermo Rauch @rauchg]

SF is the ultimate 'exchange program' of the world.

By now, it's irresponsible not to send your execs and top engineers to spend several weeks a year "studying abroad" in SF.

The data:
1️⃣ SF in Q1 2026 got 80%+ of venture dollars

[Austin @austinrath_]

@AryanBha22 It’s okay man you don’t have to censor yourself

[Austin @austinrath_]

When you ask kimi k3 to analyze your security architecture and it suddenly switches to mandarin and starts making POST requests to guangzhou

[Austin @austinrath_]

You look like a democratic socialist (derogatory)

[Austin @austinrath_]

Every guy knows that the dude who claims to be the “good guy” is always a piece of shit

That’s also why we despise Anthropic as a brand

[Austin @austinrath_]

@patinthehat0 Running the enterprise also gives you insight on the right problems to solve for with the data you collect.

[Austin @austinrath_]

@AryanBha22 @micro1_ai @RichardDreamVC @jonahvella Physical AI 👀

[Austin @austinrath_]

FWIW the most important consequence of LLMs is accelerating software development. Everything else pales in comparison. Old heads like chamath, karp and nadella actually have no idea what they're talking about. They're not in the trenches writing code with these things.

[Austin @austinrath_]

@aaronburnett Does this not also lower the cost of inference on earth?

[Austin @austinrath_]

LLMs have made software increasingly outcome driven instead of usage based. Robotics will follow the same path. It always comes back to delivering a product or service that someone is willing to pay for.

[Austin @austinrath_]

@collin_michaelm @matteo_agius @DarioAmodei They're making claude verbose so they can charge for more tokens