woeirua

I have not seen spec first development work well in real world environments at all. The problem is that you typically don't know what you don't know upfront for most product based decisions. Going back and forth in the spec to fix problems is a huge mistake, it just leads to contradictory sets of requirements for the agent.

What does work well is using AI to generate a quick and dirty prototype ala vibecoding, and then using that as a basis to regenerate the final code. Some might say it's still spec driven development, but the spec is a working prototype.

Passkeys are such a flaming disaster from a UX standpoint. I can't recommend anyone use them.

They made the same argument for gambling. "It's a victimless crime!" Except it isn't.

AI is way beyond conventional LLM architecture now. It combines LLMs with search + RL. The traditional LLM architecture hit a wall around GPT-4o. Arc AGI evals show this.

So... why would you use Fable now?

Open models wont be open for long. No one is going to release an open model capable of chaining zero-days. Even the Chinese aren't that reckless because it will just be turned around and used against them.

It'd be way better if the clusters showed the average price for all the stations underneath each cluster. Then you could get a sense of prices nationally immediately.

I thought ARC-AGI-3 was explicitly a test of raw model performance excluding the harness? Adding the harness back in doesn't tell us anything new. We've known that agents are capable of long horizon reasoning with sufficient harnesses. GPT-4(?) was capable of beating Pokemon 18 months ago but models only became capable of beating it without a harness in the last six months...?

Is there any scenario in which Github doesn't start yanking features out of the free tier to protect the core service for paying users?

Not really. Grad students are already essentially working for free.

This makes no sense. To believe this you have to think that the models are somehow being overfit explicitly on academic mathematics and it doesn’t carry over at all to more practical software engineering. I wouldn’t make that bet.

No bud, it’s just the beginning!

So… your model is 100% vibes based. Got it.

Anthropic: "Look at how dangerous our models are!"

Anthropic next week: "Why did you ban our models Mr Trump Daddy?"

A lot of people are slowly realizing that code generation was not the bottleneck in their internal development process, and waving a magic wand to make that go faster doesn't actually lead to more revenue.