Andres Vergara

@testingcatalog “No more need to carry a half-open laptop around?” I thought I was the only one doing this 😂

OpenAI is working on a dedicated setting for Codex to allow users to enable "Locked use." > Let Codex use your Mac while it's locked No more need to carry a half-open laptop around?

xAI just dropped a massive new offer for SuperGrok Heavy plan 67% off for 6 months - now $99/month instead of $300/month Includes: • Near-unlimited usage • 16x AI agents in Expert Mode • Early access to new features • Dedicated support • Access to Grok Build Early Beta • xAI’s most powerful reasoning tools And this is just the beginning There are a lot of new Grok models and features coming very soon xAI is moving insanely fast right now Subscribe while the offer is still available

We are definitely going to be paying extra, or actually, we won’t be able to use Claude anymore because we cannot afford to pay API prices for claude -p usage. $200 in API credits for claude -p = let’s move our headless workflows to codex. Mind you, a lot of us have half or the majority of our claude usage in headless workflows. A $200 subscription with no headless usage doesn’t make sense anymore.

To add some clarity: you don't pay extra. It's the same subscription, same price per month. What's new our sub now covers two separate pools: · Interactive → sub limits, unchanged · Programmatic → new $20–$200 included(!!) credit, metered at API rates

This is actually limiting usage once again. It’s steering subscribers toward interactive mode only, as if we’re meant to stay seated at the computer the entire time. For many of us, that’s not realistic — we also rely on headless long-running jobs, and this change removes that capability. Yes, there’s /goal, but assuming it fully addresses our long-running workflow needs feels a bit presumptuous. Most of us already build our own harnesses or use other headless tools. I hadn’t tried codex before the openclaw/Claude debacle, but that prompted me to give it a serious look. Now the majority of my workflow has shifted to codex. This latest change feels like the final straw. I can’t justify paying for a Max 20x subscription that’s limited to interactive sessions only. I’d rather just pay for two codex subscriptions. @AnthropicAI , please stop penalizing builders.

@noahzweben This is actually limiting usage once again. Forcing subscribers to essentially use interactive mode only is just driving people to codex where we are not limited as to how we use the subscription we are paying for.

We're launching a huge SDK credit simplification on June 15: * Agent SDK/-p draws from a new bucket instead of your interactive limits (which stay exactly the same) * Every plan comes with $20-$200 of monthly SDK credits * You can use 3P tools like T3/OC with these credits

If you think software engineering is going away, try having AI build a multi-runtime system. It breaks fast. AI can’t manage complex runtimes, integrations, and architecture on its own. You still need a human driving the architecture. AI can help. It can’t replace the human making sound decisions and stopping it from generating confident slop.

@Dimillian I can’t tell if I’m in main or a working worktree or branch on a thread. It would be nice if there was some sort of indicator on the matter. I wish there was an easy button to clean up working worktrees and branches that were just squash-merged to main from within the thread.

I’m diligently working on enhancing all the features of the Codex app to make it even better for you. If you have any suggestions or ideas for improvement, please feel free to share them with us here!

@enjojoyy I run mine in a $200 raspberry pi

Why is nobody saying that you don’t need a Mac mini to run OpenClaw? You can run it on a VM You can run it on a cloud You can run it on ANY computer You DON’T need to invest hundreds of $ just to try it out

A lot of value is still in human interaction with the client. Professionals who are good at human interaction and good with AI are the ones who stay. Others get pruned.

AI is not replacing you. Professionals who ignore it are replacing themselves.

@csallen Claud code planning mode and its execution work very very well. Much better than Cursor on a similar flow.

Someone help me understand the appeal of Claude Code, compared to Cursor: - Doesn't have a real textbox, so it's hard to select/manipulate/copy text in your prompts as you write them - Doesn't stream back AI responses, so you have to wait for them to finish before reading them - Much harder to audit all the file changes it's made - Can't go back up the tree to edit previous messages/responses - Etc. What am I missing?

Everyone is talking about how awesome 🔥 the new ChatGPT 4o image model is, and it is! But it still suffers from some common AI image generation pitfalls. Can you catch it?

🔥 I ignored all the best coding advice—and built faster than ever. Everyone says coding demands meticulous planning, structure, and precision. But the truth? I got more done last week just vibing with AI than I did last month obsessing over “perfect code.” Vibe coding flips the script. It’s not about syntax, it's about capturing a thought. A conversation. You just tell AI what you want, even vaguely, and watch it unfold. I needed a landing page, and vibe coding got it done—zero stress, pure creativity, and ready in one day instead of the many days it used to take. Here's the thing: coding can (and should!) feel as good as sketching or jamming out your favorite tune. Yes, it’s rough around the edges. Yes, it might need polishing. But for rapid prototyping, experimentation, and getting your ideas to life quickly—it's unmatched. My prediction? In five years, we'll look back wondering how we coded without vibe-driven AI. Have you experimented with vibe coding yet? What’s your take? Drop your thoughts below. 👇 #AI #vibecode #vibejam #Cursor

Me asking Cursor: Implement the following er diagram in the prisma schema. Cursor (via claude-3.7-sonnet) Sure! I'll create the additions to the schema. Done!, oh wait, looks like we also need a repository... ...and a service ...and, wait, we also need a UI for this, I'll create 7 new components for it oh, looks like we need a route for this, I'll create one ah, crap, so we also need an API for it, I'll create one, ok, now I'll fix the 67 lint errors I introduced... well, in order to fix the lint error I need to modify other parts of the prisma schema... oh wow, since I modify other parts of the schema I also need to update 4 other unrelated repositories... and, uh, their services... You would not believe what I need to update now... #AI #cursor_ai #vibecoding

Claude (and through Cursor) keeps modifying things it DOES NOT NEED TO, even when asked to not do so 🤦‍♂️ Always... Review... AI-Generated... Code. #Cursor #AI #ClaudeAI

If you're not vibe coding yet, you're are missing an opportunity wave. Throw ideas at an AI, iterate fast, and ship faster. Welcome to solopreneurship! #VibeCoding #AI #solopreneurship

AI-powered vibe coding is speeding up projects—but is it killing craftsmanship or elevating creativity? I say it's elevating creativity. What do you think? Let's talk. #VibeCoding #AI

Writing code manually in 2025? That's like mailing letters instead of texting. Let AI handle the syntax; you handle the creativity. #VibeCoding #AI #Cursor

I don't know about you, but when I use Cursor's agent to brainstorm the solution to a bug with me rather than tell it to 'fix it' and have it immediately modify code, I get much better results and can fix the bug a lot faster. and in a much cleaner way. #AI #Cursor

Amazon is launching a new hybrind, reasoning version of Nova in the summer of 2025 It's supposedly designed to offer lightning-fast responses along with comprehensive problem-solving capabilities in one package. Is anyone really using Nova?