Ștefan-Gabriel M (@legraphista) header
Ștefan-Gabriel M (@legraphista) profile picture

I build stuff, so you don't have to

Joined May 2015
Photo by Ștefan-Gabriel M (@legraphista)Photo by Ștefan-Gabriel M (@legraphista)Photo by Ștefan-Gabriel M (@legraphista)Photo by Ștefan-Gabriel M (@legraphista)Photo by Ștefan-Gabriel M (@legraphista)Photo by Ștefan-Gabriel M (@legraphista)Photo by Ștefan-Gabriel M (@legraphista)Photo by Ștefan-Gabriel M (@legraphista)Photo by Ștefan-Gabriel M (@legraphista)Photo by Ștefan-Gabriel M (@legraphista)
Ștefan-Gabriel M retweeted
Could you or your company sponsor an x86-64 server and hosting for FFmpeg’s CI? RISC-V, PowerPC or MIPS systems would also be welcome! We currently test these using QEMU on our Ampere ARM server. Please get in touch!
65
163
12
2,564
168,086
Meta built something genuinely useful. The problem is the name on the box.
10
The best open image model ever shipped. You're just not allowed to ship anything with it. Alibaba popped open the weights on a 7B image model that generates its own transparency, under a licence that bans commercial use.
1
1
26
PrismML squeezed Qwen 3.8 27B down to three values per weight. 54GB → 5.95GB. Runs on a 6GB card. They claim 98.2% of the original's benchmark scores. Their benchmark. New Tech Report 👇
1
1
42
$40M to build an AI that doesn't generate text. 24 hours later the internet had it spelling words letter by letter through a 26-option classifier, at 50 API calls a sentence. We made a 5-minute Report on it. The video wrote itself, literally.
15
Worth knowing what "zero hallucinations" means in the Jev launch: zero schema errors. The JSON is always valid. Whether the answer inside it is correct is a separate question, they report 67.8% on their own suite. Both things can be true. 100 seconds on what Jev actually does:
2
1
2
111
Credit where it's due, TypeSafe documents all of this themselves. Their launch post attaches a nuance section to every claim, including that 193.6x is "the higher end" of real-world gains. More transparency than most launches offer.
1
20
Yup, those rockets look totally the same to me 👍! Anyway, congrats to @OpenAI for #GPT6, looking forward to seeing pricing and benchmarks.
73
Thanks @dbrand, you got an awesome customer support! For my next phone, I'll definitely buy another Grip case.
Hey @dbrand, got any more of them Pixel 7 Pro Grip cases? I think mine is suffering from PTSD.
1
62
Ștefan-Gabriel M retweeted
Coming soon to starzero.ai. Shoot your scene, prompt the style. Footage courtesy of youtube.com/ManCarryingThing
2
4
156
Hey @dbrand, got any more of them Pixel 7 Pro Grip cases? I think mine is suffering from PTSD.
1
1
1
614
. @Cloudflare really looked at that subdomain and said *approved*
1
3
6
92
"I made this. All of it. You're welcome." — our starzero AI agent, after I asked it to make a video about itself I built an MCP to give our starzero Agents the same capability and asked one to highlight a day in its life.
me: "can you use whatever resources you like, and python, to generate a short 'youtube poop' video and render it using ffmpeg ? can you put more of a personal spin on it? it should express what it's like to be a LLM" claude opus 4.6:
119
Ștefan-Gabriel M retweeted
It is hard to communicate how much programming has changed due to AI in the last 2 months: not gradually and over time in the "progress as usual" way, but specifically this last December. There are a number of asterisks but imo coding agents basically didn’t work before December and basically work since - the models have significantly higher quality, long-term coherence and tenacity and they can power through large and long tasks, well past enough that it is extremely disruptive to the default programming workflow. Just to give an example, over the weekend I was building a local video analysis dashboard for the cameras of my home so I wrote: “Here is the local IP and username/password of my DGX Spark. Log in, set up ssh keys, set up vLLM, download and bench Qwen3-VL, set up a server endpoint to inference videos, a basic web ui dashboard, test everything, set it up with systemd, record memory notes for yourself and write up a markdown report for me”. The agent went off for ~30 minutes, ran into multiple issues, researched solutions online, resolved them one by one, wrote the code, tested it, debugged it, set up the services, and came back with the report and it was just done. I didn’t touch anything. All of this could easily have been a weekend project just 3 months ago but today it’s something you kick off and forget about for 30 minutes. As a result, programming is becoming unrecognizable. You’re not typing computer code into an editor like the way things were since computers were invented, that era is over. You're spinning up AI agents, giving them tasks *in English* and managing and reviewing their work in parallel. The biggest prize is in figuring out how you can keep ascending the layers of abstraction to set up long-running orchestrator Claws with all of the right tools, memory and instructions that productively manage multiple parallel Code instances for you. The leverage achievable via top tier "agentic engineering" feels very high right now. It’s not perfect, it needs high-level direction, judgement, taste, oversight, iteration and hints and ideas. It works a lot better in some scenarios than others (e.g. especially for tasks that are well-specified and where you can verify/test functionality). The key is to build intuition to decompose the task just right to hand off the parts that work and help out around the edges. But imo, this is nowhere near "business as usual" time in software.
1,589
4,638
1,113
36,748
5,180,277
This is pure gold. I've hit probably 85-90% of those issues during my time building @KamuaDotCom I've learned to never assume anything about any media file, because somewhere, someone, has some file that will break your entire pipeline and make you cry at 3 in the morning
Important post by FFmpeg developer "haasn" "Falsehoods programmers believe about [video stuff]" haasn.dev/posts/2016-12-25-f…
49
Ștefan-Gabriel M retweeted
Microsoft just released the first natively trained 1-bit model: BitNet 2B. Trained on 4 Trillion tokens. Native 1.58-bit weights and 8-bit activations (W1.58A8). Performs very close to Qwen 2.5 1.5B in benchmarks while being 1/6 of its size and twice faster.
24
149
19
1,065
108,003
Ștefan-Gabriel M retweeted
Got a picture that isn't quite right? Try our native image generation in Gemini Flash 2.0. "Can you remove the stuff on the couch?". "Can you make the curtains light green?" "Can you put a unicorn horn on the person in the green pants?" Editing in human language, not image editing tools
Gemini 2.0 Flash debuts native image gen! Create contextually relevant images, edit conversationally, and generate long text in images. All totally optimized for chat iteration. Try it in AI Studio or Gemini API. Blog: developers.googleblog.com/en…
24
61
4
761
102,395
Ștefan-Gabriel M retweeted
Open source, native audio turn detection 🎉🎉🎉 Most voice agents today do turn detection by waiting for speech pauses of a specific, short length. That's not how humans do turn detection when we talk to each other! I've been working with some friends on a new turn detection model. If you're interested in this problem or in learning more about ML engineering, come hack on a small model with us! More details below.
48
145
26
1,475
165,123
Ștefan-Gabriel M retweeted
LETS GOO! Generate full songs (4 min) with vocals in less than 10 seconds - open weights model! 🔥 It's crazy how much you can achieve from open models as of today! - going right for Suno and likes! VAE + Base model combined is < 2.5GB Open weights on the hub and a space to play around with it! 🤯
23
125
11
896
64,248