Modified Bessel Function

Honestly, the openclaw user experience is straight up dogshit. The telegram connect is bugged (I keep getting the access issue like daily), I set up reminder with cron and it never fired, and just now it fully ignored my message. Idk how it got so hyped, just not usable.

Nah it's actually unusable, it just does 3k+ tokes of reasoning without even trying anything if it's just a tiny bit unsure. Even prompting doesn't help, it just goes "user is really frustrated, I should write the fix" and never does the fix xd?

Kimi k2.6 is quite smart, but sometimes a simple "the first message in my app is not being streamed" can throw it in 160k tokens reasoning where it haven't edited a single file yet. It feels like detailed issues cause it to enter reasoning loops way too easily :(

@AdamHoltererer Use it for research-y style tasks. Like, you wanna have a feature like in other apps, but not sure on architecture/how people built that. I have special agent on opencode which takes in question -> enriches with my repo context -> I pass that to gpt pro and get in-depth answer.

I paid for ChatGPT Pro, but I don't know what I'm supposed to be using GPT 5.5 Pro for. Please, people who use it, tell me what I should be using it for, because I'm just giving everything to Codex.

@LFTH_Alex @robertrioux Ai is just some weights running on some hardware, it doesn't have a purpose, stop drinking/taking whatever you do, that would be reasonable as well.

@robertrioux The purpose of AI is to kill jobs, surveil us, and destroy the concept of truth. I think being concerned about antisocial tech oligarchs putting their greasy paws into my favorite piece of software is very reasonable.

Congratulations Twitter mob. You just killed four programmer jobs at the Blender Foundation and deprived them of 1/8th of their yearly budget. But it’s ok, now you can use Blender with a good conscience. You better send them donations now. #b3d

@arzqwarzat @atorixa00 They never ask for prescriptions for estrogel or blockers, at least in bigger cities. In the worst case you might need to try different pharmacies. Just buy online with self pickup.

@atorixa00 Я в рф скорее вообще не достану это все

1\2 от всего дела добыта, осталось добыть сам эстроген

@tomiri39 @atorixa00 For me bika used to cause elevated ALT/AST. Why cyproterone and not spironolactone would be a better question here

@atorixa00 удачи, няша! у тебя всё получится! а почему ципротерон, а не бикалутамид какой нибудь?

@ShimazuSystems You just verify that the code structure/abstractions are what you want, you check that there are enough unit tests and CI is passing, and this is fine to go unless you're working on some critical infra/data etc. You're not supposed to read all implementations...

You know at 20,487 (combined add/remove) We can assume an average and say per line this is 3-4 seconds to read, let's just boldly assume the same to understand. So what, 8 seconds per line? That's 163,896 seconds. Now let's divide that by 60, I'll round the decimal down for courtesy - that is 2,731 minutes. Now let's turn those minutes in to hours - so what, 2,731/60 again! Rounding to the nearest 1dp point that gives us 45.5 hours. Now I do work hard, so in my flow state I work for 14 hours straight. It's gunna take me 3.25 days in my state to even understand what you have produced. You do not understand, nor read your code. If this is markdown, there is zero way to verify that your AI hasn't hallucinated. This does not belong to you, it has nothing to do with you, you are a vessel. I like AI, I do not like people who pretend to be productive.

@0xSero Not "any", SGLang is a bitch on ampere. Just try run awq quants on a100, you'll see some horror. It tries to import kernels which come from vllm 10+ versions old, while installing latest, without even checking cuda/torch compatibility lmao. Not even talking how shitty docs are.

SGLang > VLLM > Exllamav3 > Llama.cpp IF you have any Nvidia hardware this is the play.

@corbin_braun Codex in vs code snce the codex app isn't stable enough for remote dev, plus some OpenCode for supplementary tasks or frontend with K2.6.

what are people using to build now?

@sircalebhammer Why even bother with that if you can get air fryer, frozen pre cooked fries, and get fresh fries at home with 0 prep time? Like I get ordering burgers and some other stuff which is hard-ish to cook, but fries or nuggets?..

Don’t do this to me. I’ve been so good.

@theo If only the remote support in t3 code was better :( For local development I prefer t3 code over codex app for sure, for actual remote, in order of stability: vs code + extension and terminal inside >>> codex app > t3 code

It's insane how quickly T3 Code has progressed. Huge shoutout to Julius and all our open source contributors. Never thought we'd get this far

@thdxr Tbh I don't think that this should even be a question. Chances are, someone who can only afford oc go sub isn't even doing anything worth privacy :D ZDR is usually a "premium" option, having this in 10$ sub is nuts and overkill, especially if that means lower usage for users.

opencode go is currently zero data retention however we can increase limits and make it all more sustainable if we collect data to train future open source models you can opt out of this - is that something you'd be ok with?

@kr0der At my work we use self-written OpenCode orchestrator, with some custom pipelines and agents, so it can actually write and run smoke tests on new code instead of plain CI. Based on internal tests this is way better than just making agent read diffs

we cancelled greptile recently due to price increases, so we've cut down to Bugbot + Codex reviews now also finding Bugbot’s resolved rate to be ~75% which is pretty good - i wonder how it’s calculated what PR review tools are you guys having success with?

@thsottiaux Terminal panes, so we can read/use 2 terminals at once. Also, there is some weird rendering going on, where it breaks TUIs, and also in remote it's not really stable, I've seen my terminals randomly close way too many times, to the point where I have to run terminals separately:(

Send us feature requests for codex in the form of an images 2.0 generated image. It makes it easier for codex to implement if we decide to go for it. Saw some good ones today already that codex is cooking on.

@CookingRobotGuy @FactoryAI 1) M2.7 is a big upgrade over M2.5 2) it's great for agentic tasks, I don't make it write code for me - for that I have codex (on my main codebase even opus isn't reliable), but review-verify loop, where I use pipeline like on the screen - M2.7 does it job quite well.

@theio666 @FactoryAI Yo minimax 2.7 is NOT reliable for code . At a minimum you MUST ask "are you certain? did you trace through all the code?" to get even 60-80% accuracy. but of course, depends on how complicated the code is. But that's my experience since minimax 2.5 days. K2.6 is much better.

Which model reviews code best? We benchmarked 13 models on AI code review across real PRs and the results are surprising. Spending more tokens did not result in better code review. A $1.25/PR model beat another that was more than 2x the cost. Meanwhile, budget models at $0.15/PR delivered ~80% of the quality of frontier models while being 10-30x cheaper. In fact, cost only explained ~21% of the difference in code review quality.

@svgmzr @itcaat Generate a high level review with info on what new modules/abstractions were added, check if code lives in sensible places, etc. You don't have to carefully read 10k loc diff, but you can still, well, vibe-review it :D

@itcaat Так смысл вайбкодинга в том, чтобы никогда не смотреть в код. Как можно поревьюить 10KLOC?

- я тут навайбкодил, можешь пулик поревьювить? - пошел нахуй

@WillHavePeace @EstieMaddie "Turn off your brain and be happy" is a decent strategy for life, but some people strive to be better than that ;)

@EstieMaddie Happy people are Republicans. If you want to be happy, then become a Republican. Pick a career that makes money Follow a budget. Workout. Go to church. Get married. Vote for mass deportations. Have kids. And you'll be happy too

🤩 THIS IS AWESOME!! 🤷🏻‍♀️ “DO YOU THINK THEY ARE DEMOCRAT OR REPUBLICAN?”

👀 finally it looks usable :D My small customizable OpenCode-gitlab integration pipeline creator I've been working for some time :) Still needs quite a lot of polishing, but happy with the progress.