@lochan_twt Good riddance. That website was a cesspool of egomaniacal losers. I'm glad it's gone.
English
Theres_can_be_only_one
469 posts
Theres_can_be_only_one
@wait_theres_mor
Your mother's house. Katılım Ocak 2018
163 Takip Edilen23 Takipçiler
@0xSero Do you mean Coding Agent Harness or General Agent Harness(everything else, plus coding) ?
English
Yeah, but that’s kind of the point... it’s a benchmark meant for "benchmarking".
They’re not just measuring whether the model can do the task, but how it performs under strict constraints. Humans could recreate something like this without the internet too... That’s not really in question.
What actually matters is the time and efficiency. LLMs are useful for coding because they can often do the same work as humans, just faster. If it took an LLM 10 years to recreate something like FFmpeg, then it wouldn’t be any better than a human in practice.
That doesn’t make LLMs useless, but as a benchmark, this is about comparison. Tests like this give insight into how close these systems really are to human-level capability. or even AGI.
English
@deedydas true. i guess my point is that the capability this benchmark measures may not be that critical
most of the time our agents can access Internet when they work, and they should. a model that’s great at doing things from training data may not actually be great at real world tasks
English
@pvncher Openai should give Dario a raise, he's the responsible for this.
English
@Tiny_Fish Will you guys collect the search queries for training
English
Starting today, TinyFish Web Search and Fetch are free.
For every dev and agent. Across the galaxy.
No credit card. Generous rate limits.
Grab your API key from tinyurl.com/bdzer5x6
May the web be with you.
English
@UnslothAI Are you guys really still talking about openclaw in may 2026? Do you live under a rock or something?
Hermes Agent my dudes, keep up.
English
We made a guide on how to run open LLMs in Claude Code, Codex and OpenClaw.
Use Gemma 4 and Qwen3.6 GGUFs for local agentic coding on 24GB RAM
Run with self-healing tool calls, code execution, web search via the Unsloth API endpoint and llama.cpp
Guide: unsloth.ai/docs/basics/api
English
@chkdg8 @Techjunkie_Aman Well... It’s open source. Why don’t you make your own version instead of demanding that people on the internet work for free… you worthless leech.
English
@Techjunkie_Aman Too bad the developer is a crybaby when it comes to Linux.
English
People don’t realize how insane this is.
DuckStation just pushed a release with 100+ fixes touching almost every layer:
Rendering:
• GPU pipeline fixes
• texture handling improvements
• shader parsing and post-processing fixes
Core emulation:
• CPU timing corrections
• interrupt handling fixes
• DMA and memory accuracy improvements
System stability:
• VRAM corruption edge cases resolved
• audio backend consistency fixes
• UI and fullscreen transition fixes
Impact:
• more accurate PS1 emulation
• better frame pacing
• reduced glitches and crashes
This is not a small update.
This is continuous engine-level refinement.
Techjunkie Aman@Techjunkie_Aman
DuckStation's Android version is effectively discontinued. The last update was released on April 25, 2025. The app is still available on Google Play and works fine on most devices, but the developer no longer maintains the Android version. So if issues appear: • No updates • No bug fixes • No support The developer explicitly states that emails asking for help will be ignored. System requirements: • ARMv7 / AArch64 / x86_64 • 64-bit recommended • ~1.5GHz CPU for 32-bit Even today, it remains one of the best PlayStation 1 emulators on Android.
English
Boston Dynamics just had a brutal C-suite shakeup
CEO Robert Playter retired in February, the COO and CSO bounced right after, CTO Aaron Saunders jumped ship to Google DeepMind… and a ton of senior engineers and researchers are gone too.
This is happening right as the company gears up for IPO and tries to scale Atlas from flashy lab prototypes into real mass production.
Hyundai (the parent company) is cracking the whip: they want tens of thousands of these humanoids in their factories ASAP.
Basically going from “cool robot video dreamers” to “actual factory workforce suppliers” overnight. Board is in full panic mode. With Optimus, Figure,Unitree and the rest breathing down their necks, losing this much talent at crunch time… oof.
English
@NVIDIA_AI_PC 40 items on my model folder so far... Yeah I like to experiment with them a lot!
English
Be honest — how many local models do you have downloaded right now? 👀
English
@TechByTaraa Well, only grifters and tech-illiterate normies used it. Nvidia and OpenAI praised it because it burned ungodly amounts of tokens while doing absolutely nothing.
It’s still shit to this day. People now have far better options, so of course it has fallen into irrelevance.
English
@jun_song You don't need a local LLM if you're a tech-illiterate fuck who just wants to "vibe code" single-user apps.
"Claude, make a Twitter clone, acquire the first 1 billion users, and make no mistakes."
English
You don’t need Local LLM if you:
> don’t care about your data and privacy.
> don’t mind frontier labs cutting your usage on every update.
> okay with them banning your account for no reason, and not replying on customer service.
English
Theres_can_be_only_one retweetledi
@rovarma The biggest problem with Linux adoption was never that Windows was too good. It’s that some people in the "Linux community" are completely detached from reality.
English
This is 100% true, but Linux devs were and are in denial about this.
There’s a reason the install page for every cross platform app looks like
* Download Windows installer
* Download OSX package
* Click here for Linux install instructions that may or may not work on your distro
gingerBill@TheGingerBill
I don't know if a lot of people have thought why this happened. To make Linux viable for the layman, Valve had to make Proton (derived from Wine) so that Win32 API became the first and only stable ABI on Linux. Why did Linux Distro devs not care about stable ABI historically?
English
Have you ever built anything useful in your entire life?
Or are you just a stupid retard? People like you are the worst. Your intelligence is so limited that you’re below a small 4B AI model it knows more about code, math, science, and the world in general than you, you worthless jealous uneducated retard.
English
300 ms latency rollback local test for 64, seems promising :3
English
You stupid tech-illiterate retard demands that people work for free on the internet.
Why don't you take some responsibility and start sending some patches?
Why don't you take some responsibility and start making donations so competent people can send patches?
Why don't you shut the fuck up and stop being a stupid retard online?
English
your product is critical to a lot of applications
and you have this sort of response?
you need to wake up, stop acting like a spoiled 3 year old child, shoulder the responsibility you have to fix the issues that exist.
whether you wanted that responsibility or not, you have it because your product is so integral to so many things
if you don't want that, sell the product to someone that will take that responsibility seriously
FFmpeg@FFmpeg
We get this a lot and it's a non-sequitur. Many people run marathons or 10Ks or have hobbies which don't pay the bills but consume a considerable amount of time. But it gives them a sense of achievement and being part of something bigger.
English
@TeksEdge @ElevenLabs Yes, and that's a good thing.
Competition drives innovation!
English
Oh nice voice cloning on xAI console is here. Should @ElevenLabs be worried?
TestingCatalog News 🗞@testingcatalog
XAI 🚨: Voice cloning is now available on xAI Console in the US. > Create a custom voice in less than 2 minutes or select from our library of 80+ voices across 28 languages to personalize your voice agents, audiobooks, video game characters, and more. This also means we will see custom voices on Grok soon. I hope they won't be restricted to the US only.
English
We implemented @karpathy 's MicroGPT fully on FPGA fabric.
No GPU.
No PyTorch.
No CPU inference loop.
Just a transformer burned into hardware, generating 50,000+ tokens/sec.
The model is small, but the idea is not: inference does not have to live only in software 👇
English
@sudoingX It's not even close. Hermes is the an absolute beast.
English
@ahmetkoc @sudoingX @sama @AndrewCurran_ Because he doesn't own a non profit company called OPEN something you fucking retard.
English
@sudoingX @sama @AndrewCurran_ why dont you ask elon musk to opensource all tesla and speacex codebases ?
English
Elon's going to show up uninvited to this GPT-5.5 party like the witch in sleeping beauty and deliver a powerful curse.
English