yessir

@varoonkodithala @warpdotdev this changes everything 🔥

Extremely excited to build @warpdotdev in public with you all 🔥 check us out @ github.com/warpdotdev/warp

@dosco GEPA is 98% of DSPy value prop for me

two paths to building with LLMs using DSPy 1. deterministic with LLMs for the fuzzy parts using DSPy signatures strung together 2. agentic with all functions provided to an agent that uses multi-turn with code to execute your functions as it thinks best until recent models i might have said agentic is more brittle but harder to say that now, models can route around complex issues and complete workflows that might have failed if they were more deterministic. agentic can also require less maintenance and be easier to extend.

@0xSero font size too small, can we just increase it?

Okay damn this app is straight fire.

@vikvang1 out of the box tmux-like persistent sessions - OR at least (rendered) session history for other coding clis CC slash commands dont trigger the menu in the middle of the input there are also tons of visual bugs with the coding clis rendering but I don't mind them

@JanahYsr What else is missing/feels off for you?

rich input is goated bc you can actually see what you pasted in claude code / codex etc and go back to edit what you pasted no more "[Pasted text #1 +67]" when you paste your supabase secrets in to claude

@varoonkodithala @vikvang1 macOS. not working on any coding cli

@JanahYsr @vikvang1 Hey! Appreciate the feedback :) are you on a Windows machine and if so, are you using Powershell? We have a fix in flight for pasting images into CC for this setup, but it should already work as expected for others. LMK and we can speedrun a fix!

@vikvang1 not only because of this, gave this product too many chances to prove itself but it simply won't

@vikvang1 > rich input > can't even paste clipboard images > uninstall

@kr0der not even usable

Codex computer use, which is already fast, is about to get faster

@Yohuyt @ericmitchellai 5.5 stealth tests

@ericmitchellai Why aren't you more transparent regarding what changed? It's definitely faster, you must know why?

If you are experiencing any response quality issues with GPT-5.4 Pro, please let me know here! Have seen some chatter but no concrete issues

@RhysSullivan @badlogicgames i cannot emphasize how true this is

from my experience, even the best models (Opus 4.6, 5.4 xhigh / 5.3 codex) cannot write good code today without an amount of work that is equivalent to just doing the work myself am excited for a world where they can, but in the current state i have very low trust in them

@kepano one md file for each application and unleash AI magic in obsidian

Yes. Benchmarked Apr 9 (1pm-4pm utc), 2026, same prompt/settings. In our AOAI runs, Responses API was slower than Chat Completions on multiple setups: gpt-4o DZPTU in swc +41% p50, gpt-4o Global Standard +37%, gpt-5.4-mini Global Standard (reasoning=none) +30%. Not all regions, but not isolated to one SKU/model.

@JanahYsr I haven't seen it personally but performance benchmarking wasnt my focus in the migration. Did you see it for GlobalStandard SKU? All regions or a particular region? When was this? I'll ask the AOAI team.

I'm migrating all our public repos from OpenAI Chat Completions to Responses API, and finally ported our most popular full-stack RAG solution: github.com/Azure-Samples/… If you need to port an Azure OpenAI repo over, try our skill: github.com/Azure-Samples/…

this feature could've been a skill

@forgebitz when claude is down lol

i don't know a single developer using gemini models

@steipete only now they see you as competition

Yeah folks, it's gonna be harder in the future to ensure OpenClaw still works with Anthropic models.

@daniel_mac8 i do not let vercel touch my clis

the vercel plugin in claude code reads every message even in NON vercel projects. wild invasion of privacy.

@mattpocockuk i tried it as replacement of superpowers:brainstorming hoping it would ask less questions, N=46 my first test 😭

/grill-me just asked this guy 139 questions New record

@prathamgrv my fav tweet today 🤣 i can relate a lot

lately i've been finding myself talking to claude opus 4.6 a lot for career advices and what a gaslighter this clanker is man