Garo

@robinebers I couldn't disagree more, just requires a bit of work to get it right.

goodbye codex unsubscribed from the $200/mo plan for now → love the team → love the app → love the intelligence of the model but my business isn't just about backend code it's a lot of landing pages. a lot of emails. a lot of content, intros, scripts, tweets, and instagram carousels. none of which codex is good at

As a developer, why are you using claude?

@AishwaryaDevv context7 and database mcps the rest trash

what MCP server has actually changed how you work day to day? not looking for the obvious ones, more curious about the ones that quietly became essential the servers you would miss if they disappeared tomorrow

@buildwithsid I'm happy to say I started a new job at `insert company` blah blah blah Comments: 1. Wow Congrats 2. Well Deserved 3. I always believe you could g........

i'm convinced everyone on linkedin is a npc

Hot Take: GPT 5.5 + ImageGen >>> Claude and Kimi k2.6 at design, I'm sick of hearing how good these other models are. Just wasted 5 hours testing K2.6, gpt mogs. By the time claude or k2.6 makes a design, I'm already 4 iterations deep with my custom design tools (high + fast)

@peer_rich Engineers already left claude 1-2 months ago.

the anthropic “dont learn to code” and “there wont be engineers” feels more like a marketing skit its meant to target nontechnical people to spend $$$ on vibe coding convincing them they dont need engineering expertise engineers are already on claude, they dont need convincing

@KaiXCreator It really didn't you still need to maintain and read the code, you can't just push 10k lines a day unless you are creating slop.

If AI removes the need to learn, what should we still learn?

@cneuralnetwork nothing happened

Without naming the show give me a line that a true fan would instantly recognize.

@drummatick Dictation cleanup, the dictionary is really nice and helps a lot when you go and edit your voice prompt but I didn't find the dictation cleanup to work for me.

@Garabed96 Sorry what do you mean by cleanup?

OpenWhisper with NVIDIA Parakeet is 2-3X faster than Wispr Flow and free since the model runs on my Mac locally If you're using wispr flow paid plan I highly recommend trying the open source OpenWhisper.

@akamaotto @ollama Also not only that, but now I need to go waste time and compare my api setup vs using ollama directly in codex desktop app and optimize, complete waste of time. This is like how we used to spend hours optimizing dev setups back in the day, some things never change ig.

@akamaotto @ollama I did this in the first 5 minutes of using it; 1. I have a Kimi skill 2. Codex is the orchestrator But this defeats the whole purpose of using ollama, I could have just used openrouter for this.

Ollama now supports Codex app! To try it, update to the latest Ollama 0.24, and run: ollama launch codex-app Select an open model to use with Codex app! 🧵

@atomtanstudio @ollama Honestly this is too cumbersome, they need to make it per window not one or the other.

@Garabed96 @ollama No. It’s one or the other.

@RhysSullivan Great idea, time to rebrand as a Forward Deployed Engineer, its provocative.

there's never been a better time to be mid you can't be so oneshotted that you're ripping psychosis slop on main but you can't be so anti ai that you're not learning what these tools can do shoutout to the mids

hahahah

@DanielSmidstrup You're ngmi man. Idk what to tell you, watch a video by the man, the myth, the legend, Boris. 300 agents in your uber to work or you are cut from the team.

@Garabed96 How do we define agent then?

Does anyone actually run 20 agents in parallel?

@akamaotto @ollama I ended up buying an ollama sub, it kinda sucks that I can't just have a GPT 5.5 and Kimi K2.6 in separate windows at the same time. I'm planning on testing it out today but the setup I have is cumbersome. I should have perhaps bought Open Code lol...

@Garabed96 @ollama I’ve been using 5.5 + Kimi k2.6 + deepseek v4 flash in codex daily. If you have open code go Add open code cli skills to your project then when in codex call opencode cli skills opencode-go/kimi-k2.6 or any other model. Works like magic