bhaktatejas922

The shoggoth knows

factual

Its true - Cursor, Cline, and many others still use xml for tool calls. In JSON, the model needs to "focus" on escaping characters correctly while also sampling from a reduced token distribu…

"Creating models that learn developer biases" has a great ring to it - maybe we should make that our mission statement. Thats exactly what we're doing with our models. The Next-Edit com…

nice catch. the html example was using a hardcoded snippet we forgot to uncomment. fixed

or lets be right faster

yeah its such a better experience! - it was really the difference maker between the cursor experience and everything else previously

yeah speed and flow state are for sure linked. People love to say Cursor is a Claude wrapper but they miss the reality that Cursor is a fast apply wrapper. Intensely sticky user experience

working on it

yep!

so frustrating how slow edits are in Q dev

how often do you run into merge conflicts?

sounds like flow state to me

it's a bit unclear why a model works best here. in short - smart diffing is edge case hell and you'll never capture all of them

yeah same. I feel like Opus tends to be slightly more sycophancy leaning on technical topics

you could make the argument about any startup really. To me its the same reason they don't build the foundational model for legal, for sales, etc.. - everything comes at a cost. Allocating resea…

yeah even claude is well over 11% error rates with search and replace

done! Yeah we have ZDR options as well, just email us to enable it info@morphllm.com Morph via OpenRouter is always zero data retention

I've never had so much fun coding in my life - you should definitely give it a try again!

No, we use autoregressive llms. Diffusion models would be super interesting here. Mercury is doing some interesting work with diffusion in code gen but still too early to tell if it'll get good e…

the power of hacker news! New models are listed there now

key word incremental - for fast apply to be useful it should be so fast and accurate that most people don't realize there's a model there at all

I think an MCP would do the job. We're shipping one out as we speak

and speculative edits is faster

Correct, but morph is a LLM as well. In practice its basically Big LLM using small LLM as a tool call

exactly. The point is that none of the users even realize a model is doing the apply - it should be so accurate and fast that it feels like its not there

the v2 model listed currently points to morph-v3-large. We're working with them to get v3-large and v3-fast listed

hm maybe but correction/issue detection is a much harder task for models. If you pipe back the errors in it could work, but personally still see Fast Apply as the better approach

Wow that was fast - this is awesome. it shouldnt be a problem unless your code has both <code> and <update> internally. 1 or the other should be fine

Like most things its a tradeoff. Developer tolerance for errors is extremely low - but the error rate for Fast Apply is even lower