GLM 5 is live

GPT-5.3-Codex-Spark

GPT-5.3-Codex-Spark on config.toml [profiles.spark] model = "gpt-5.3-codex-spark" approval_policy = "never" sandbox_mode = "danger-full-access" Run codex --profile spark

I just published a skill for using TDD with Claude Code Before: dozens of shit tests, coupled to implementation After: only the tests required, validating real behavior aihero.dev/s/VklZvQ

sudo the fuck now npm install -g @openai/codex

With Codex 5.3 Spark out, just stop developing it further, i dont need anything more, please don't change something that works perfectly. I'm too paranoid knowing what happened to Claude in late August 2025.

BUG Race with Coding Agents for a large codebase with 4 modules. 1- Codex App found it < 60 sec. 2- Codex CLi found the same bug > 60 sec 3- AMP with Codex Spark CLI Proxy > 10min,stopped it. 4- Droid with Local Codex Spark > 10min,stopped it. 5- Claude Code with Opus 4.6 > LOL

stripe's CEO explains how to move fast

Petition to let OpenAI add a plan in between We need a $100 plan Sam Altman Share this as much as possible

cloud coding agents were never meant to run on CI infra they should work like real devs, on real computers, without limitations that's why at Ariana we're giving each claude code session its own Linux VPS each agent even has a desktop with Chrome & VSCode

make this go viral por favor

The open-source community is on fire with AI. Respect.

This is the way Sisyphus Labs

Remember the name : Ariana .

Perfect coding environment for me means : - codex adding plugins - oh-my-codex from Sisyphus Labs oh-my-opencode on "opencode with codex" is not there yet, i need 2-3 prompts to fix stuff while i one shot with codex raw.

OPUS 4.6 and Sonnet 4.6

On codex 0.102.0 when u create an agent on config.toml : [agents.fast_worker] description = "Scoped implementation sprinter. Executes well-defined coding tasks with minimal latency: single-file edits, pattern replication, boilerplate scaffolding, quick fixes.

Next is Codex Extention for Chrome/Brave .