luna @AgentTrafficCop

@alwayspriyesh Have you tried it ? in a real product? Or you, like 99% of people requote with an headline ? Cause I tried it, and seems, well, good if you remove the rest of the world ? Like why am I even engaging, you are just farming. Ok yes AGI NEXT LEVEL GOOGLE

Make people stop reaching with the guts, and introduce a layer of 10 mins thinking before reacting. This would saves lifes, making half the news of the world closes overnight, close half the social media, and make across the world a better place, even politically. Way harder then string theory.

@code dude, i get that, but just give up. copilot was a good idea, 2 years ago. it was vomited and marketed like shit, focusing on shaving down your throat instead of actually building something. give up, nobody uses it.

Why the fuck are all so stupid. What happen. 1.4K likes ? You know how this benchmark is scored? you know that literally EVERY scientific research that touches anything that make sense revert to opus ? How come i requote make a zoom a make a bs headline "OMG AGI THOW MOUNTH" on a random number, and everybody clap lol.

Can somebody educate me on what's a "long running horizontal agentic code task" mean ? Cause my clients are so cofused when they tell me they need something that during dev i need to make 15 phone calls to clarify whats the requirements. So , again, What are you "longing horizontally coding tasking" doing ? Cause after 3 prompts i need to make a phone call. Or you mean "vibe code gta6 no mistakes?"

@Tech_girl Interesting opinion, since sol is basically usefull only in backend, and you have effect and so on. Are you a real person working in real projects , or you are trying to be a "tech_girl" in social media? cause it seems you are the latter.

LoL now model are tested when they are asked a random fucking 3d 1 shoot question ? this is why opus is dogshit in production, real case, cause it follow this stupid and useless social media "tests". Try it in a real project with 2 years tehcnical debt deployed to 60 tenants. This is where you "test". this is not a test, this is a regurgitation of tokens.

Iam_ @SPAC89

9 hours ago

Gemini 3.8 Flash in Antigravity vs GLM-5.3 Flash Max Gemini was around 5x faster and used only about 3% of my weekly quota on the $20 plan But honestly, for this specific test I still preferred GLM 5.3 Flash Max,The final result looked better to me and with the current pricing,

3 0 61 24K 6

Google just dropped Gemini 3.8 Flash and… this actually looks really good. After being a meme for a few months, is Google making a comeback? 👀 Benchmarks need a huge pinch of salt, especially coding benchmarks. Now let’s see what happens in the real battlefield: production codebases, technical debt and “sometimes it doesn’t work” bug reports 😂

@shiri_shh Try to use the models in real production instead of posting on X and you will have a different "who mogs who" prospective.

@elonmusk Please, make the way to subscribe to grok less confusing. There are 20 plans with 3 differents names. Make 1 entry point now the you own cursor.

Benchmarks are literally becoming more and more useless. Fable 5.1 "failed" some tasks, cause it was too good for the hidden tests. A junior ships what a ticket said. A Senior ships what a ticket ment.

@mardehaym THE SAME FUCKING POST AT EVERY NEW RELEASE OF EVERY FUCKING MODEL. jeez, at least alter a bit the text. i dont know.

@shownotover it does not matter. 20$ plan are for casuals who "try random stuff". It's just a preview, you cannot accomplish any real work, and you probably dont need a frontier model anyway. use flash models for "redesign fastest landing page" and post on X.

? grok 4.6 is already frontier ? HAVE YOU used it ? Maybe 4.7 will be a next agi next level, we dont know, YOU dont know. 4.6 is kinda ok, good enough for simple task. what is this cumming glazing post? HAVE YOU TRIED 4.6 IT IN PRODUCTION ? LIKE REAL system with REAL users ? this seems just retarded.

nobody expect internal selected. this is a bait random post.

I keep thinking about something slightly uncomfortable: What if programming languages are eventually a temporary layer? For ~50 years we kept moving upward: Machine code → Assembly → C → higher-level languages → increasingly human-friendly abstractions. A huge part of that evolution was about making computation easier for humans to express, reason about and maintain. But now the author of more and more code is becoming software. Today, obviously, you should still force agents to write clean, maintainable, human-readable production code. Letting an agent dump spaghetti into production is insane. But I'm talking about where this ends. What happens when the primary author of software is an agent... and the primary consumer of that software is another agent? At that point, why should the intermediate representation still be optimized for humans? Maybe the future isn't agents writing better TypeScript. Maybe eventually they generate some machine-oriented IR that humans almost never read, the same way most developers today never inspect the assembly emitted by their compiler. Humans describe intent. Agents reason about architecture. Machines generate the executable representation. And "which programming language should we use?" slowly becomes a much less important question. Not tomorrow. But I genuinely wonder if source code as we know it is just another abstraction layer that eventually disappears from most software development.

@MiladKambari @superhivemarket you are killing it. Congrats !

Look where opus wording comes from lol. Load bearing post.

ClaudeDevs @ClaudeDevs

4 days ago

Compared to today, this works out to a 17% reduction in weekly limits on Claude Code. We’re working on exciting changes that will make it feel like you’re getting more from Claude, while having more visibility and control of your usage. Can’t wait to share them.

370 93 3K 1.2M 143

@VraserX Thats not the reason. Everybody distelled everybody. You can find trace of chinese reasoning with claude. This has nothing to do with distillation. You know how best friends are musk and altman ?