And from a sales perspective, that’s probably not wrong.

I think a new (or newly critical) engineering challenge is how to build in a way that keeps a codebase that’s churning out features from crumbling under the weight.

I’m not convinced we could train easily here, or that it’s worth the investment compared to (previously) spending fractional cents on Luna, or now paying even less on Jev. Especially given that these numbers are not meaningful to our margins.

I have a dozen different things at work that are currently using LLMs as classifiers for different questions. I don’t have the time, data, or resources to fine tune a model for each of them.

I haven’t had a chance to plug in Jev yet (waiting on approvals), but if it has the general intelligence claimed in the press release, then Laya is in no way comparable for my use case, and whatever TypeSafe has done is a substantial innovation over the Laya paper.

I’m not saying it was right or moral, just probably inevitable.

That said, when I’m reducing a list, I still use reduce.

Dario already showed there was an appetite for this.

Obviously insurance companies make money on average for every policy. If they didn’t, they would go bankrupt. So on average, every time you buy insurance, you lose.

So insure only what’s required by law, or things that would throw your life off track if they vanished.

We can round up all the bored teenagers we want, but it’s not putting the genie back. Better start adjusting our systems to account for it.

Maybe we’re Notion/Evernote? Buy Hackpad, plow a ton of money into Paper (which was legitimately good), then quietly deprioritize it.

Maybe we’re actually some kind of enterprise document productivity suite? Buy HelloSign. Plow a bunch of money into a desktop app. Pull more plugs.

A lot of smart people were trying. We made a lot of bets (too many?). None of them proved to be a second act, and the competitors eventually caught up.

Wouldn’t surprise me if the providers just remove that ability and lock the model once the conversation starts.

Not sure about the OAi Pro plan, doesn’t look like the 80% Luna price slash made its way into the quota system.

You could also try tuning down the effort level on Opus. It makes a huge difference in token consumption and you might be able to get away with lower than you’ve set

Mills did ramp back up, but it’s unclear to me if they used it as a chance to do so slowly/preserve margins. Lumber never got close to pre-pandemic levels.

Tariffs probably also play a role here. About a quarter of US lumber comes from Canada, and barbed wire is just steel with a little bit of processing.

For products with little value add there’s not anywhere for the tax to be absorbed, and no real way for domestic producers to quickly scale up, even if they wanted to.

I’m not sure if they let you skip the cache write cost on the first turn. That would imply cross-user caching infrastructure or special casing the default system prompt to give you a discount. Maybe? Away from the computer but you could try a “hello” in a fresh session and see what was billed.

You should use --disallowed-tools to prune any tools not needed for the task. Note that this is also a perpetual game of whack-a-mole since they’re always adding new tools.

Any kind of scrolling back, copying text, using their menu system - basically anything that isn’t typing characters has had/still has unaddressed bugs.

OpenAI shipped a competitive model and I’m over in Codex now. I have yet to hit a bug.

If you’re holding the SOTA crown, people will put up with your buggy mess. As soon as that crown slips your pile of trash becomes a huge liability.

They have one of the more compelling cases for rolling their own.