RJ

So, you seem not to understand the costs involved with tokens. And that's ok. Not many people do. Currently, Tokens are highly subsidized. You're being ratelimited for the exact thing they are trying to avoid. If your company, "BridgeMind" is so good, I sugguest that you invest some time in running your own model, and discover for yourself just how expensive it is to "run X windows" This is the "build or buy" decision you should have learned in school.

This is what my workflow looks like now. 5 Codex terminals running GPT 5.4 in BridgeSpace. Only 1 Claude Code terminal. I used to run 7 Claude Code agents with Claude Opus 4.6. That was two weeks ago. Now I can barely keep one running. Codex with GPT 5.4 Medium Fast is stable. Available. No rate limit surprises. No outages. I never thought I'd say this but Codex is a better daily driver than Claude Code right now. Not because the model is better. Because it actually works. Cancellation confirmed. Moving on.

I'm still bullish on AI, but people need to stop treating it like it's going to save them. You can only save yourself. AI isn't some magic silver bullet. It has limits.

@maxcagr @gianfrancopiana With test suites, you can measure lots of things. Could be response time for a headless browser, could be total test time. Heck, I've tried measuring round-trip times to the database to see if there were any issues there. Lots of things to measure in a test suite.

@gianfrancopiana “You give it a command that measures something” what did you measure?

@milan_milanovic More so.

When Opus 5.0 with 10M context is released, do you still have a job as a developer ?

@MaMoMVPY This was stated a year ago; Dario is often wrong on his timelines.

I must say I am increasingly suspicious about the comments from the AI bosses - why do they need to make these completely over the top and so obviously unfounded predictions about how AI (or rather LLM) will impact economic development? To me it is an indication of something not being quite right in their own business models - they will need to attract more and more investors as they are likely going to face a funding squeeze sooner rather than later. And again let me stress - LLMs are great and can surely increase productivity, but presently AI companies like OpenAI and Anthropic are making heavy losses. This means sooner or later LLM prices need to go up, and potentially a lot. And what happens then with the case for LLMs? Might it be that the entry-level lawyer might be both a lot better AND cheaper than an AI "agent" that is priced at what is needed to make Anthropic or OpenAI profitable?

@socialwithaayan 48 markdown files

🚨 BREAKING: Someone turned Claude Code into a full game development studio with 48 AI agents. Claude Code Game Studios mirrors an actual studio hierarchy inside your terminal: → 48 specialized AI agents each with a distinct role → Art Director, Level Designer, QA Lead, Sound Designer and more → 36 workflow skills covering the full game development lifecycle → A coordination system that manages agents like a real studio pipeline → Build complete games with AI agents handling every department 3,000 stars and rising fast. 100% free and open source.

@GoogleResearch Middle out????

Introducing TurboQuant: Our new compression algorithm that reduces LLM key-value cache memory by at least 6x and delivers up to 8x speedup, all with zero accuracy loss, redefining AI efficiency. Read the blog to learn how it achieves these results: goo.gle/4bsq2qI

I think this is the best post I've ever made.

@nateberkopec "Ruby is slow"

Replaced a `preload` with a single query that returns a list of ids instead, then we use Ruby to filter that list.

@Urban_Cipher @Polymarket +50% of your labor is supposed to be spent on AI tokens

@Polymarket Isnt this company supposed make workforce to 50% due to AI ?

BREAKING: OpenAI reportedly plans to nearly double its workforce to 8,000 ‌from 4,500 by the end of the year.

@mkobach 6mo

Is AI moving so fast that learning anything AI right now will be obsolete in 18 months?

You win some, you lose some

@Rahll I said the same thing dude. MFW: The guys selling AI sell more AI.

Wait wait wait waaait a second.... He's saying he wants his $500,000 engineer to actually cost him $750,000 by using $250,000 worth of AI tokens. Wasn't AI supposed to make things cheaper, not cost 50% more?

Let me check the math, bro... $500K engineer + $250K in tokens = $750K engineer. If that's your moat, you just made SWEs more expensive. Not more replaceable.

RIP #ChuckNorris

@robbyrussell Son of anton, strikes again

Claude Code just intentionally DELETED nearly 550k LOC from one of our apps!?!

Pokemon Go's 'Scan a PokeStop' feature secretly trained AI sold to third parties, all without user consent. The data collected was far more than just for gameplay improvement. #PokemonGo #DataPrivacy #AI

Your gaming data is being collected and used without you knowing. It's valuable labor for AI training and billion-dollar projects. Are you aware of how your playtime is being monetized? #DataPrivacy #Gaming #AI

Harness engineering offers high ROI in AI now, but it's a depreciating asset. Future AI models will likely absorb current scaffolding, making today's efforts obsolete. #AI #Engineering

Pokemon Go players are unknowingly training AI. By playing and scanning PokeStops, they provide the labor and data that powers these systems, making players the actual product. #PokemonGo #AI #Data