You can do that yourself with opencode and a $10-15 subscription to a good model.
I used to be a skeptic, but this month the pairing of tooling and models I discovered is finally good enough to build a static Go app with a TypeScript front-end, along with all the deployment and compile steps, despite knowing nothing about either of these.
Insufficient or bad wrapper will hobble a model. An insufficient model will mean the wrapper cannot help much. Both have to be in place, and this month they finally came together for me to create ready-to-go stuff.
Heads up that this is "more true" for non-reasoning LLMs. Reasoning gives an LLM a lot more runway to respond to out of distribution inputs by devoting more compute on understanding the code it's changing, and to play with ideas on how to change it before it commits.
1. You can run decent local AI now - see /r/LocalLlama. You pay the electricity cost and hardware capex (which isn't that expensive for smaller models).
2. Chinese APIs like Moonshot and DeepSeek have extremely cheap pricing, with optional subscriptions that will grant you a fixed number of requests of any context size for under $10 a month. Claude Code is the bourgeois option, GLM-4.7 does quite well on vibe coding and is extremely cheap.
Local LLMs that you can run on consumer hardware don't really do anything though. They are amusing, maybe you could use them for basic text search, but they don't have any real knowledge like the hosted ones do.
Gemma 3 27B, some smaller models in the 8-16B size range, and up to 32B can be run on hardware that fits in the "consumer" bracket. RAM is more expensive now, but most people can afford a machine with 32GB and maybe a small graphics card.
Small models don't have as much world knowledge as very large models (proprietary or open source ones), but it's not always needed. They still can do a lot of stuff. OCR and image captioning, tagging, following well-defined instructions, general chat, some coding, are all things local models do pretty well.
What I meant was that you cannot just clip one in your pocket and use it as your daily driver. Try signing a receipt, or even saying to someone (as for example I did, this past Tuesday), hand me a piece of scrap paper and I will write down the instructions for processing acrylic paint waste water.
You may want to use the new "derestricted" variants of gpt-oss. While the ostensible goal of these variants is to de-censor them, it ends up removing the models' obsession with policy and wasting thinking tokens that could be used towards actually reasoning through a problem.
Great advice. Have you observed any other differences? I’ve been wondering if there are any specialized variants yet of GPT-OSS models yet that outperform on specific tasks (similar to the countless Llama 3 variants we’ve seen).
I used to be a skeptic, but this month the pairing of tooling and models I discovered is finally good enough to build a static Go app with a TypeScript front-end, along with all the deployment and compile steps, despite knowing nothing about either of these.
Insufficient or bad wrapper will hobble a model. An insufficient model will mean the wrapper cannot help much. Both have to be in place, and this month they finally came together for me to create ready-to-go stuff.
reply