samtheprogram

This is so, so wrong, at least in the context that you're pitching it.

Fallout is great example of a subset of their IP making a huge turnaround. Halo is not.

To sell all the IP at once is grave mistake. They would easily make more on separate deals. Netflix not taking it all at 800m seems like a mistake.

> I suspect that many of the loudest voices pushing back against AI coding tools never practiced programming as a hobby or saw it as especially fun, and the fact their job is changing won't make them start now.

Never thought Justin Searls would say that. So I reread the whole comment over and over again, and it clicked.

I liked building and making stuff work. I also didn't mind and often enjoyed coordinating with (good) clients, managers, or other engineers. I didn't love being a manager and requested a few years ago to go back to an IC role.

That said, I think I didn't like being a manager because of the turnaround time and/or quality of people I managed (partially due to my blind spots in managerial skills, at least in some cases). I improved but it wasn't worth the headache for me at the time.

With agents, I feel some of that pain again, but it also fixes some of the problems I had with it; relatively consistent quality and I find that I can lower my bar of communication with an LLM (I'm not writing in public to a real colleague) while simultaneously getting a lot more out of it, and it tends to not misconstrue what I've written.

I keep going back and forth between loving it and pining for being an old fashioned IC.

All that said, I really enjoy the puzzle of programming, in ways that I couldn't really always do at work -- "it's good enough", etc... I've done it in free time this year but it's hard not to view it as a waste of time now.

Not sure where I'm going with this besides expressing my current thoughts and somewhat agreement. /rant

To be fair, I'm reading it as satire of AI. So not liking it is kind of the point.

It's a release candidate. That's what they are for, to catch bugs. Nothing to see here.

...except the interesting tidbits shared like this sibling comment: https://news.ycombinator.com/item?id=50142208

Pro-tip: it costs very, very few tokens to have an agent (harness) sleep a random amount of hours between each toggle. Just tell it to. It's not burning tokens on `sleep`, just to call it.

It might not hit input cache at that rate but better than a block or ban. Keep the context window small.

If everyone else you know has JavaScript turned on, congrats, you're still fingerprinted.

It's almost like C has keywords and pragmas for your compiler too. Neat!

It's really not different. This is just backward compatible (e.g. for intellisense). Honestly better than a #define that I have to declare to get a build going in a lesser C compiler (worst case, looking at you Apple).

'use cache' works at the function level as well and is very handy to reduce scaling bottlenecks in a pinch.

> putting occupants at risk in seconds

Yeah, uh, I totally agree with you and won't be reading the rest.

> The most successful businesses are those that deeply understand something about their customers, about how to sell to their buyers. The most has never mattered. An army of agents doing endless busywork is not the path to success.

You're talking about the best businesses. Arguably the article is talking about the mean of the "startup" tranche of businesses.

Google isn't a SaaS. That's (is or is tangential to) the kind of businesses this article is referring to. And they make decent money, they just aren't a unicorn that didn't die.

DS v4.1 Flash is roughly equivalent. It's going to get some things right/better that GLM flash doesnt and vice versa.

Reminds me of YouOS. RIP

Could be Intel specific -- the writer has a touch bar.

It doesn't seem much worse to me if at all. Not a noticeable difference. That said, the laptop is only a year and a half old, with the battery cycle count at 314.

I also wonder if I'll have a "oh, the battery life _is_ much worse" moment soon, but not yet!

Yet to be determined if they actually ship or do stuff at a reasonable rate. It only has any notoriety because of the subject matter and jumping on the hate-DHH train.

I guess they've been "laying the groundwork" of a fork that claims to just want to rename things and do little else for over a month?

In the current/legacy market, people accept bugs in software because software was expensive and the alternatives are often few for that same reason.

Tomorrow's software will need to be _perfect_, or else it won't differentiate from "Claude, build a basecamp-style project management tool for my team".

Probably the same reason as oh-my-zsh, you don't need 90% of it. Further compounding the problem in an agent harness is that you are polluting the context window by throwing the kitchen sink at it.

Unfortunately I don't know yet because I updated for Xcode 27 + iPhone dev, so having an iPhone plugged in 24/7 is heavily spewing the results towards "much worse".

If I can still reply mid-to-late next week I'll let you know.

Per the article, that's the price OpenAI is willing to pay for an exploit that covers any account or integration one connects to their OpenAI account. Let that sink in.

I don't think the other commenters mentioning how server-side vulnerabilities aren't as lucrative in the black market are making that connection.

Maybe macOS 26 was lot bigger... but I went straight from Sequoia to Golden Gate and freed up like 30 GBs. I didn't have enough disk space to download the update prior, so the extra 30 GB wasn't that... but could've been caches I guess. I assumed it was dropping Intel?

They already knew to drain the basin. Control-F in a PDF is way less annoying than watching a 9 minute long (for the algorithm) YT video.

Sure, you can search the video transcript... but all the YT video is doing is regurgitating the manual.

Did you not catch the emphasis? They did end up writing about it.

Sure, not for the same reason, but... got'em.

Tailwind is so pervasive it's 100% in pretraining _and_ RLHF.

I've done several Tailwind projects and rarely did models from various providers need to reference Tailwind documentation.

> go after profitable niches and avoid unwinnable fights

If you know Apple's history, this should stand out to you.

I desperately miss their old XULRunner app. No longer worked through the API as of ~3 years ago and was a sad day for me.

He didn't say 90% of the coding jobs. He said LLMs would write 90% of the code. As in be LLM generated.

If you've ever been confused about which turn is which with top down navigation, then this is for you. While driving a car, being able to at a glance know exactly what turn is what would have saved me countless wrong turns, my parents probably doubly so.

One classic: there's a fork in the road on the way to my parents. One side is "straight", the other side is an exit. I took the wrong, exit side of that several times while my app said to go straight before I just knew the route and didn't need GPS at all. Top-down isn't as good in these cases because of densely packed roads and back to back exits.

That said, it's "good enough" and I doubt this takes off.

I had to login the other day to use it. Probably getting slowly rolled out.

If you just want it to generate UI out of nothing, the benchmarks aren't really for that.

If you want to generate a UI based on specific user input of some kind, then they are.

I'd suggest using one model for UI and another model for tacking onto that UI. LLMs are great at pattern matching, and benchmarks don't really capture one-shotting desirable UI.

That said, benchmaxxing is a thing and your experience with models is a thing. Benchmarks are fuzzy and should be taken with a grain of salt.

> designed

What?