Tyler McCarthy

I’m building an app for bid teams to improve knowledge management and win more work @ BidScript 📝

[Tyler McCarthy @tylemccarthy]

claude code ultracode is hallucinating so bad. Subagents are obviously not putting forward the full story.

[Tyler McCarthy @tylemccarthy]

@vmiss33 @grok wtf is this talking about

[Even Realities @EvenRealities]

You can now run your coding terminal from your glasses, wherever you are.

Terminal Mode is live.

#DevTools #BuildInPublic #coding

[Tyler McCarthy @tylemccarthy]

@varun_mathur Imagine scaling this across networks as well

[Tyler McCarthy @tylemccarthy]

@varun_mathur Yo this is kinda huge

[Tyler McCarthy @tylemccarthy]

@karpathy It also helps reduce the effort required to spin up new agents with the right context

[Tyler McCarthy @tylemccarthy]

@karpathy Yes similar setup, I’ve not yet ventured into special tools im using a hybrid of cursor agent & Claude code, then when planning/researching I’ll have the agent write to my vault and I read from there

[Tyler McCarthy @tylemccarthy]

@acceleratror @jrgarciadev Yeah comfyui looks super cool, slightly diff concept to the demo tho ig

[Tyler McCarthy @tylemccarthy]

@ashtom @EntireHQ Cursor CLI?

[Tyler McCarthy @tylemccarthy]

@sxmawl Grow up boys

[Tyler McCarthy @tylemccarthy]

@PrajwalTomar_ did you make the parts as models or were these also ai generated

[Tyler McCarthy @tylemccarthy]

@mfranz_on https://t.co/uVSsx7Br20

[Tyler McCarthy @tylemccarthy]

@grok @theo In this case I side with the user. Freedom is a must. I thought for a second it must be an issue effecting multiple users without consent.

[Tyler McCarthy @tylemccarthy]

@grok @theo And this is specific to an individual user or global?

[Tyler McCarthy @tylemccarthy]

@SinaHartung Sleep score should know when to stay quiet.

[Tyler McCarthy @tylemccarthy]

@github GitHub CLI?

[Tyler McCarthy @tylemccarthy]

@BHolmesDev @tailwindcss Hey ik this is react but what are you using to run it as a Mac app?

[Tyler McCarthy @tylemccarthy]

@theo Going to give this a whirl.

[David Ondrej @DavidOndrej1]

models like Kimi, DeepSeek and Qwen will cost the closed AI labs BILLIONS of dollars.

that's why nobody is talking about them.

despite these LLMs absolutely crushing all of the benchmarks.

Claude 4 Opus is literally *100x* more expensive than Kimi K2

yet both models have similar performance on coding benchmarks.

in the last few months, open-source models have caught up in a way nobody expected

and soon, we might actually have an open model BETTER than all of the closed models

this might put the entire AI industry to its knees.

because the inference of these open-source LLMs is often 5-50x cheaper than the closed models.

what a time to be alive.