I've seen people post this same advice and I agree with you that it works but you would think they would absorb this common strategy and integrate it as part of the underlying product at this point...
The people who build the models don't understand how to use the models. It's like asking people who design CPUs to build data-centers.
I've interviewed with three tier one AI labs and _no-one_ I talked to had any idea where the business value of their models came in.
Meanwhile Chinese labs are releasing open source models that do what you need. At this point I've build local agentic tools that are better than anything Claude and OAI have as paid offerings, including the $2,000 tier.
Of course they cost between a few dollars to a few hundred dollars per query so until hardware gets better they will stay happily behind corporate moats and be used by the people blessed to burn money like paper.
> The people who build the models don't understand how to use the models. It's like asking people who design CPUs to build data-centers.
This doesn't match the sentiment on hackernews and elsewhere that claude code is the superior agentic coding tool, as it's developed by one of the AI labs, instead of a developer tool company.
You don't see better ones from code tooling companies because the economics don't work out. No one is going to pay $1,000 for a two line change on a 500,000k line code base after waiting four hours.
LLMs today the equivalent of a 4bit ALU without memory being sold as a fully functional personal computer. And like ALUs today, you will need _thousands_ of LLMs to get anything useful done, also like ALUs in 1950 we're a long way off from a personal computer being possible.
That's $500k/yr, and I guarantee there's a non-zero amount of humans out there doing exactly that and getting paid that much, because of course we know that lines of code is a dumbass metric and the problem with large mature codebases is that because they're so large and mature, making changes is very difficult, especially when trying to fix hairy customer bugs in code that has a lot of interactions.
Doesn't specifically seem to jive with the claim Anthropic made where they were worried about Claude Code being their secret sauce, leaving them unsure whether to publicly release it. (I know some skeptical about that claim.)
A lot of it is integrated into the product at this point. If you have a particularly tricky bug, you can just tell Claude "I have this bug. I expected output 'foo' and got output 'bar'. What went wrong?" It will inspect the code and sometimes suggest a fix. If you run it and it still doesn't work, you can say "Nope, still not working", and Claude will add debug output to the whole program, tell you to run it again, and paste the debug output back into the console. Then it will use your example to write tests, and run against them.
I learned about JQBX and similar platforms through people that reached out as I've been sharing Jukebox around and they seem like they were beautiful corners of the internet.
We're building Zigpoll (https://www.zigpoll.com), a survey platform focused on zero-party data collection — think post-purchase attribution, customer feedback, and segmentation — all done directly on your site without relying on third-party cookies or offsite links.
We initially built it for Shopify, but now it’s fully embeddable, supports headless implementations, and integrates with tools like Klaviyo, Zapier, n8n, and Snowflake. One thing we’re especially proud of is how fast and unobtrusive it is: polls load async, don’t block rendering, and are optimized for mobile and low-latency responses.
From a tech angle:
Frontend is all React, optionally SSR-safe.
Backend is Node.js + Postgres, with a heavy focus on queueing + caching for real-time response pipelines.
API-first design (public API just launched: apidocs.zigpoll.com).
We recently open-sourced our n8n integration too.
If you're a dev working on ecom, SaaS, or even internal tooling and need a non-annoying way to collect structured feedback, happy to chat or get you set up. Feedback welcome — especially critical stuff. Always looking to improve.
[1] Surveys. Thinking about how to tighten up the onboarding experience, improve brand awareness, improve in-app data analysis, and how to integrate AI in new and exciting ways... and handling customer support tickets!
I've been working on Zigpoll as a one-man project for a while now: https://www.zigpoll.com/ it has traction and solid growth (~100% YoY for the past 3 years) but the larger the numbers the harder it gets to double each year.
In a past life I would have thought this would be the easy part given the product market fit but it's hard to figure out growth channels that are scalable and cost-effective at this stage. Burning what would otherwise be a large salary month on month in search of growth is mentally taxing when it doesn't deliver. Metrics across the board only seem to tell part of the story so it's tricky to figure out what needs changing and what's worth doubling down on.
If anyone has experience doing this sort of thing - please get in touch!
Yeah, I love this stuff! Compiling data from multiple pages into a single paragraph in the time it takes to read one page? Great stuff. I can't imagine living without Perplexity.
Oh, sure, it hallucinates a lot, and in dangerous ways, but even if I have to manually corroborate all the citations, I'm still saving time, especially insofar as it reveals whether or not I'm barking, broadly, up the wrong tree.
It's especially good for comparisons, because the results of two disparate search terms can be collated into the results.
Could this be done without LLMs, but only vector embeddings? Hm, maybe. Algolia is maybe the 80 for 20, but does Algolia have a web index?
Not when it's code that was only hard to write because you needed to know the right incantations to pipe data between different services.
Now you see the incantations that mostly work and the job of transforming it is easy.
Java's Bouncy Castle crypto library is a good example of this. The thing you're trying to do might be simple, but to do it, you might need to instantiate 8+ Java classes. It doesn't mean it's complex to read or hard to debug.
> The thing you're trying to do might be simple, but to do it, you might need to instantiate 8+ Java classes. It doesn't mean it's complex to read or hard to debug.
I’m skeptical that code that needs to instantiate eight separate classes will remain easy to debug in the general case.
LLMs give you a lot of false confidence, just because something looks right doesn't mean it is.
Especially with cryptography you should NEVER use LLMs. Read the docs, write down some notes, and make sure you properly understand everything before you use it. You need to really think it through before you end up leaking user data or worse.
Not 100% on the nose with what you're looking for but I built Zigpoll (form builder for on-site surveys and forms) that may useful: https://www.zigpoll.com/
