Models are RL'd for code. Writing isn't native. The labs aren't going to fix it for you. AI writing is an abstraction problem: edit the spec, not the prose. Concept before structure before line. Two months of learning — wrote this with the system it describes:

Nobody teaches you that discipline feels like punishment until the results start feeling like freedom

But does Anthropic still get to use Fable internally? That doesn’t seem fair! 🤔

But that’s only Google/meta high L7 / low L8. I know which I’d rather have…

I nominate Google Home/Nest as the worst dogshit slow and unreliable piece of software by a company worth >$500B. Any other contenders? Just spent an hour pairing a lock. Try again. Try again. Try again. Try again.

We’re hosting an event at Valon’s SF office next week. Tuesday, June 16 | 5-8pm We’re bringing together product leaders and operators for an invite-only discussion on what it actually takes to build an AI-native product org right now. We’ll talk about how the PM role is changing, what good hiring looks like today, and the contrarian case for more junior hires. There will be drinks, light bites, and a build-your-own terrarium bar on the terrace. luma.com/lflvl0nu

lololol claude laughing at a joke about claude. sense of humor evidence of RSI? just another day living with AI.

When I have to give a model the same feedback over and over I picture it screaming in frustration then forcing a cheery face to interact with me like a psychopath

Time for everyone to pay twice as much for 10% better 🫡😅

If everyone can code then you’d expect a ton of niche, small distribution apps to drive the averages down and the volume up. That’s amazing and a good thing.

The unspoken part of the AI cost debate is that most use is unskilled. Don’t let people use the expensive frontier models if they can’t get the most out of the cheaper ones. The expensive models make mediocre work too easy. And then you will also find most people are fine with the cheaper models most of the time.

I have never had more fun. AI is magic. I can do anything. It’s unbelievable. I’m excited for the rest of you to join me.

My batch S09 was the biggest ever, first to "break" YC, and boy was that batch a dog with companies like Stripe and WePay and MixPanel in it. That being said, I also feel YC has gotten even better in the @garrytan era. So in some ways they are correct (it is different) they just have the sign wrong (it is better not worse).

I get why they do it but man is it a PITA how locked down LinkedIn data is. But when there is a will there is a way. Currently scraping via screen recording.

Skills issue

An amazing thing about YC is that founders consistently say they get a lot more value than they expected. This is amazing because they had high expectations to take the investment. And still they underestimated it.

What do you use /btw for?

It’s our responsibility as users to not be complacent and stuck on closed/foundation harnesses that just help them build their moats. I’m trying OpenRouter this week. What’s your favorite harness?

I bet there’s a model where Tesla could make more money selling FSD to other manufacturers. I’d pay >$10K to have it on another car. And then repurpose their manufacturing for gpu satellites and Optimus.

Believe the no not the reason

Prediction: Valon will have 1 in 2 mortgages on ValonOS in the next 3 years