I tried for years to make handwriting on the iPad work, but it never clicked for me. I randomly encountered a reMarkable Paper Pro Move recently, instantly purchased it. For the first time in my life, I no longer tote a paper notebook with me. The software, simplicity, low-latency, form factor, writing feel, everything is absolutely perfect for me.
Claude Code does a lot regarding optimizing context usage, tool output, sub-agent interactions, context compaction, and stuff like that. I don't imagine OpenCode has the same financial incentive to decrease the token cost Anthropic takes on under the subscriptions.
Claude Code does a lot of work in optimizing context usage, how much output is included by tools and how that's done, and when to compact. This very well may make the cost of providing the subscription lower to Anthropic when Claude Code is used. It's well within the realm of possibility if not likelihood that other tools don't have the same incentive to optimize the buffet usage.
Not sure where that goes in the analogies here but maybe something about smaller plates.
Wow, thanks everyone. First HN post ever and it’s this intentionally terrible experiment that I thought was the dumbest weekend project I ever did, and it hit the front page. Perfect.
I’ve been reading through all the comments and the range of responses is really great and I'm so thankful for everyone to take the time to comment... from from “this is completely impractical” to “but what if we cached the generated code?” to “why would anyone want non-deterministic behavior?” All valid! Though I think some folks are critiquing this as if I was trying to build something production-ready, when really I was trying to build something that would break in instructive ways.
Like, the whole point was to eliminate ALL the normal architectural layers... routes, controllers, business logic, everything, and see what happens. What happens is: it’s slow, expensive, and inconsistent. But it also works, which is the weird part. The LLM designed reasonable database schemas on first request, generated working forms from nothing but URL paths, returned proper JSON from API endpoints. It just took forever to do it. I kept the implementation pure on purpose because I wanted to see the raw capabilities and limitations without any optimizations hiding the problems.
And honestly? I came away thinking this is closer to viable than it should be. Not viable TODAY. Today it’s ridiculous. But the trajectory is interesting. I think we’re going to look back at this moment and realize we were closer to a real shift than we thought. Or maybe not! Maybe code wins forever. Either way, it was a fun weekend. If anyone wants to discuss this or work on projects that respond faster than 30 seconds per request, I’m available for full stack staff engineer or tech co-founder work: [email protected] or x.com/samrolken
This was an unserious experiment meant to illustrate the gap and bottlenecks that are still there. I agree that there's a lot that could be done to optimize this kind of approach. But even if you did, I'm not sure the results would be viable and I'm pretty sure classic coding (with LLM assistance and all) would still outperform such a product.
As an unserious experiment, I deliberately left this undefined for max hallucinations chaos. But in practice you could easily add the schemata for stuff in the application-defining prompt. Not that I’m saying that makes this approach any more practical…
No, I wouldn’t say that my hypothesis is that non-deterministic behavior is good. It’s an undesirable side effect and illustrates the gap we have between now and the coming post-code world.
I thought about doing that, or having the LLM create and save HTML components, but for this particular experiment I wanted to keep it as pure and unfiltered as possible.
As soon as they put a persistent Salesforce brand banner across the top which did nothing but waste space and put that ugly logo in our face every day, my team started our transition off Heroku pretty much right away.
I might consider something like this, but I can connect four monitors to my MacBook Pro. How easy is it to find a Linux-friendly laptop which can also do this?
I have a lot of luck using 4o to build and iterate on context and then carry that into o1. I’ll ask 4o to break down concepts, make outlines, identify missing information and think of more angles and options. Then at the end, switch on o1 which can use all that context.
I think it’s possible that it can be a useful step. Fleshing out vocabulary can help with all the real-world use, because you’ll have more context and can fill in meanings, and get more out of real world use. But probably only as a booster to real world use. By itself, it does very little.
Why must LLMs or “AI” beat or match the smartest and most capable humans to be considered to solve a real problem? There’s been a lot of technology invented and in widespread use that solves real problems without having human-like intelligence.
https://x.com/samrolken