Mati
@iagooar
Followers
354
Following
12K
Media
181
Statuses
2K
Name is Mati - MAH-tee. Founder & chief builder @ https://t.co/L8Z829072P. Co-founder @ Podigee. AI, Codex, agents, content, podcasts. Christian, dad, husband.
PL
Joined November 2024
I can run 10 Codex CLI instances for the price of one Codex GUI. And the CLI feels closer to where the true power of agents is: coding and scripting. Codex GUI will still have its own fans though, so not a bad call - and it shows how @OpenAIDevs and @OpenAI is committed to devs
1
0
0
The future of support is a tightly integrated self-improving loop. Humans will still focus on decision-making and coordination, but most daily operations will happen automatically.
We've reached an all-time high of 87.6% autonomous resolution rate on @vercel support cases. Best part: people truly love it. Even when the AI can't help, the overall UX is better (we auto-fill the ticket form). Last week I had my "CEO supports day". It's now clear to me that:
0
0
0
I've been trying out Codex vs Claude with many "side-by-side" tasks while working on @BullBarAI There is clearly not ONE winner as the models have vastly different strengths and weaknesses. Codex + Rust is the most powerful, precise combo when it comes to infra and backend,
0
0
0
This is why I am building @BullBarAI We need to move fast, but we cannot risk giving agents access to our entire digital life.
sometime this year we’re going to get the clawdbot moment but 100x crazier with full computer use. not sure the world is ready for that.
0
0
0
I have been YOLO'ing AI agents for too long, and I always wanted a remote coding env so I can check progress from the bathroom. That's why I built @BullBarAI, and the first beta is coming really soon.
AI agents are powerful. Running them on your laptop with your SSH keys, AWS credentials, and personal data is risky. Setting up a proper isolated environment is a pain. That's why we're building BullBar!
1
0
1
Codex 5.2 xhigh is just RAW POWER you can use. The harness plays a supporting role.
@DanielMiessler I honestly think that Anthropic scaffolding is mostly compensatory due to their having an inferior model. GPT Codex 5.2 works better out of the box than Opus with scaffolding.
0
0
0
Aha! Found a piece of pre 1.0 final Rust code written by myself in 2015 👴🏻💥
0
0
1
I cannot tell if Codex 5.1 max or Opus 4.5 wins. Guess they are just good enough. Getting annoying though, I have to jump between the two if one struggles to solve a given problem. Should be easier to work with models across different providers. They should work as a team.
1
0
2
Opus 4.5 is SOLID It keeps being the best option for UI / frontend for sure. On the backend, it is fast and reliable. GPT-5.1-Codex-Max has a real contender.
0
0
0