Kairo

oh-my-opencode v3 just mass-produced senior engineers basically a coding team that never sleeps tried it and in 1 hour it implemented functionality that would take days

@Marsmensch @opencode Thanks for sharing this, super interesting setup. Keep us updated with insights as you tweak it further

I spent way too long optimizing my @opencode + oh-my-opencode model config in the past days to find out IF I could do without Anthropic. Turns out I can! Codex, GLM & MiniMax are perfectly balanced now for my use cases and I will cut down one more subscription next. Full breakdown to save you some time

@bcherny Why does it burn so many tokens compared to cli

A massive ship from the Claude Code Desktop team. We've been dogfooding this for a while, and can't wait for you to try it.

Anthropic just raised $30B at a $380B valuation. that’s serious capital especially with Opus 4.6 pushing hard into enterprise and coding workflows they’re also leaning into the safety + regulation positioning and talking about covering infrastructure externalities capability race is obvious but the governance narrative is getting louder too does funding scale equal model leadership long term or just longer runway?

GPT-4o is officially retired from ChatGPT today usage numbers were tiny but the emotional attachment clearly wasn’t a lot of people liked the tone the "warm"vibe the specific personality GPT-5.2 fully takes over now interesting how fast we get attached to tools that are only a year old did 4o feel different to you, or was it mostly nostalgia?

GPT-5.3-Codex-Spark dropped yesterday. smaller, speed-focused optimized for near real-time coding loops when latency drops enough, the interaction model changes less waiting more iterative steering the question is whether ultra-fast models become the default and larger ones become escalation paths for daily dev work, what matters more right now: raw capability or responsiveness?

xAI is in transition re-org team changes deeper SpaceX integration meanwhile Grok’s roadmap keeps expanding voice, coding, video, automation ambition is high but frontier models need cohesion as much as speed what matters more right now for frontier labs: move fast, or stay stable?

Claude Opus 4.6 is leaning hard into multi-agent systems task decomposition shared context parallel reasoning across agents less chatbot more distributed worker model enterprise usage seems to be accelerating are multi-agent setups actually useful in daily work yet or still mostly impressive demos?

@slow_developer codex 5.3 high just feels more usable day to day, faster, cheaper on tokens, less agent babysitting opus is strong but after 5.3 it’s hard to justify going back

claude code might be 3-4x faster but i barely use it because codex 5.3 high does the job better most of the time and why sit around waiting for an agent to build a feature? even if you wait, it’s still 10–100x faster than us 2025 was the year writing code fundamentally changed

GPT-5.3-Codex feels like a real step forward stronger coding benchmarks more agent-style execution better handling of long, structured tasks it’s clearly built for workflows, not just prompt demos at the same time, ads are being tested in free tiers while infrastructure spending keeps climbing frontier performance isn’t cheap if ads become standard in AI tools, does that change how you use them?

@ibuildthecloud hard disagree, less time on boilerplate and yak shaving means more free time for actual thinking, if it’s 1.1x you’re probably fighting it not steering it

AI coding sucks. 10x the code, 10x the problems, 1.1x the productivity.

@synthwavedd what’s the actual issue you’re seeing, hallucinations, reasoning or just worse outputs overall? Havent tried it in a while

insane just how bad gemini 3 pro has become. 3 flash is better half the time nowadays 🤦🏻‍♂️

@auralix4 try claude code, first time i ran it was the first time i actually heard my mac fans spin up

@EXM7777 still too expensive for what it does, burning tokens on reasoning instead of shipping code

Opus 4.6 is overthinking too much imo

@ivanfioravanti mega release weeks are fun, but real signal is evals + pricing, glm-5 looks solid though

Fasten your seatbelts! Mega AI Release week!

@thdxr @pierbapt @rauchg @vercel wait what you worked at vercel?? had no idea, just assumed you were building your own things this whole time

@pierbapt @rauchg @vercel i was actually the one who built it

Vercel Domains is so great. Pick a domain. Buy it. Connect it to a project in 10sec. Done. insane work @rauchg / @vercel

@thdxr codex 5.3 is noticeably faster and way more token-efficient than before, so for me that efficiency + lower cost makes it a no-brainer for dev workflows, not looking back at opus after that

codex is by far a better coding model than opus - anyone who knows anything understands this but the whole industry should reflect on why opus is the most popular people assume whatever is the smartest will win but the old rules of product are still what determine everything

AI isn’t staying purely digital robotics, embodied systems, real-world autonomy less demo energy, more deployment signals once intelligence leaves the browser the risk profile changes this phase feels slower but more consequential

big tech is pouring massive capital into AI infrastructure this year the buildout is obvious data centers, chips, compute everywhere at the same time, smaller sectors are already feeling pressure tools that compress hours of work into minutes infrastructure winners are clear the downstream impact is less predictable

@Siddcodes1 why steal other posts, i have seen it already multiple times from others....

What will happen to my data on google drive if I cancel the extra storage subscription ? Will it just vanish..??

@melvynx 3h on Opus? that’s gotta be a spicy invoice, especially if it’s chewing tokens the whole time

> Fix tests. Make no mistake 3 hours and 28 minutes later... AI will get depressed after this task.