Thank you for the kind words on the rules/documentation! It was definitely an iterative process to figure out how to get good results.
We have an llms.txt and llms-full.txt (~9k lines) which contains all our documentation. Feeding these to the claude didn't get great results, it was just too much information.
We manually compressed our llms-full.txt into a rules file (~1.5k lines) which declared the API upfront and provided snippets of how to do different things with callouts to common examples. This condensed version did better but would cause Claude to make subtle mistakes.
Looking at the kind of mistakes Claude made, it seemed like a human could make those mistakes too (very useful feedback for us to improve our API ). We thought “what's one of the smallest fully contained examples we can make that packs a bunch of info on how to use Instant?” That would probably be useful for both a human and an agent. And indeed it seemed to be the case.
> Looking at the kind of mistakes Claude made, it seemed like a human could make those mistakes too (very useful feedback for us to improve our API ).
This is something we've found for our API -- just having LLMs attempt to use it helps us identify things that we haven't documented well or placed enough emphasis on (for things that are critical but are non-obvious or may be drowned out by other less important information). Improvements that help the LLM tend to be good for developers too.
Life imitates art: afar you’re describing there is basically _The Secret_ (I.e. if I wish hard enough for something then eventually it will come true), except it’s LLMs that get wish-fulfilment, not us.
Huh, you've reframed The Secret for me - now I see it being, if you wish hard enough for something then eventually the Universe will make it so, just to shut you up.
InstantDB (YC S22) | Founding Engineers | ONSITE - San Francisco | Full-time | https://instantdb.com
We're looking for talented individuals interested in solving some of the hardest problems in real-time databases and front-end technology. If you love building delightful developer experiences and want to work on cutting-edge tech, we want to hear from you!
Our current stack: Typescript + Clojure + AWS Aurora Postgres
* We offer client SDKs, a CLI tool, and web GUI to interact with Instant. All written in Typescript
* We have a server written in Clojure for managing websocket connections, parsing queries, running permissions, and broadcasting novelty.
* All of which sits on top AWS Aurora Postgres
Open Roles:
* Typescript Engineer - Help build the best type system for our database and improve developer experience.
* Full-Stack Engineer - Build frontend sync engines and UIs that developers use for hours.
* Backend Engineer - Solve hard database problems and build infrastructure to handle 100k+ connections.
Additional details:
* Location: We're based in San Francisco, CA!
* In-person or open to relocation only: We're a small team and we really do prefer all working together in person!
* Compensation: Sliding scale between 0.5%-2% equity and 150k - 200k base + medical/dental/vision benefits
Our vision is to be the infrastructure for all apps of the future. If this jives with you we should really talk . Send us an email: founders@instantdb.com with a bit about yourself, and a project you've worked on.
We offer a generous free tier which doesn't limit your number of projects, never pauses, and available for commercial use. The pro plan is $30/mo and then pay for usage.
