LlamaCoder just got a lot better with "high quality" mode! It uses two chained LLM calls (1st call to make a plan, 2nd call to code) to create much better apps.

This is the first of many improvements I'm making! If you want to see the PR including the prompts I used for this new coding agent, take a look here:

AppKit is the full-stack toolkit to build onchain app UX 🪄 ✅ Social, Email, and Wallet Login ✅ Embedded Wallets ✅ Crypto Swaps ✅ On-ramp Integrate with just 20 lines of code across 10+ languages for all EVM chains and Solana. Onboard millions of users for free today.

nice, also try out gemini-2.0-flash-thinking for making apps here in a similar way:

nice, does pretty well!

Reminds me a bit how perplexity works?

Directionally yes! I suspect Perplexity is doing a lot more than 2 calls in the background. The secret sauce to these LLMs is breaking down the problem and doing a lot of LLM calls, then aggregating them :)

You guys are killing it! I just discovered your services yesterday and will be reaching out for a call to discuss a few questions I have about inference scaling for multiple customers.

Thank you, great to hear!

When deepseek v3?

Soon! Working on it

Tried it, really helps the quality, good job

Great to hear!