Levente Orban (@orbanlevi) header
Levente Orban (@orbanlevi) profile picture

cloud & infra and tech | swimmer & runner

Hungary
Joined November 2010
Photo by Levente Orban (@orbanlevi)Photo by Levente Orban (@orbanlevi)Photo by Levente Orban (@orbanlevi)Photo by Levente Orban (@orbanlevi)Photo by Levente Orban (@orbanlevi)Photo by Levente Orban (@orbanlevi)Photo by Levente Orban (@orbanlevi)Photo by Levente Orban (@orbanlevi)Photo by Levente Orban (@orbanlevi)Photo by Levente Orban (@orbanlevi)
GrokBot is overrated. Can't do easy tasks.
1
45
OK, I've stopped using @SparkMailApp - full of AI bullshit, doesn't work correctly, always messes up email sending, and has a lot of other user experience issues. When I use a mail client, I don't want to talk to fucking noob AI agents - I just want to send emails, that's all.
3
2
184
I've completed AGI Strategy with BlueDot Impact! I have mapped technical AI trends to future capabilities, developed threat models through kill chain analysis, applied defence-in-depth frameworks to design protective interventions. bluedot.org/certification?id…
1
67
Apps are becoming less important. Data are becoming the real product. Of course, apps will still exist for special use cases. But for many everyday actions, the app itself is not the most important part anymore.
1
37
Tested the 35B MoE on Nvidia Spark GB10. Here are the test results: polaroi8d.github.io/puding/
Aloha! 🌺 Meet Ornith-1.0, a family of open-source LLMs specialized for agentic coding. Ornith-1.0 spans the full parameter sizes including 9B Dense, 31B Dense, 35B MoE, and 397B MoE. It achieves state-of-the-art performance among open-source models of comparable size on coding benchmarks including: ✅Terminal-Bench 2.1(77.5) ✅SWE-Bench(82.4 on verified, 62.2 on pro, 78.9 on Multilingual) ✅NL2Repo(48.2) ✅SWE Atlas(41.2 on QnA, 42.6 RF, 39.1 TW) ✅ClawEval(77.1) Post-trained on top of gemma4 and qwen3.5, Ornith-1.0 employs a novel self-improving training strategy in which reinforcement learning is used to generate not only solution rollouts, but also the task-specific scaffolds that drive those rollouts. By jointly optimizing the scaffold and the resulting solution, the model generate higher-quality solutions in agentic coding.😎 All models are released under the MIT license, enabling full commercial and research use. 📖Tech Blog: deep-reinforce.com/ornith_1_… 🤗Huggingface: huggingface.co/collections/d…
58
Confirm you're human on Instagram.
1
44
terminal driven everything
30
Are you here in Amsterdam for #kubecon?
1
49
-7C es a csavo rovidgatyaba kint kavezuk az asztalnal a kavezo elott 😂
34
ChatGPT why love '—' the sign?
64
Everyone keeps saying AI is bullshit, but can you actually do tasks faster with it or not? If not, then don’t use it others will. It’s that simple.
60
Last Saturday, mesh. hosted the first meetup in Szeged for builders 🧑‍💻, programmers, and everyone who enthusiastic about technology. Big thumbs up to the mesh. and thanks to Guild for their office.
110
I’m beyond proud to share something very close to my heart: my amazing wife, Gréta Orbán-Gergely, has been nominated for the Young Hotelier of the Year 2025 Award. %
1
51
Seeing her recognized for all that effort is truly inspiring, and I couldn’t be more proud If you’d like to support her on this journey, your vote would mean the world to both of us: 👉 lnkd.in/dCGAmZ8M %
1
30
Please don’t forget to confirm your vote via the link sent to your email afterwards. Thank you so much for helping me cheer her on - every vote is a little boost of love and support that she truly deserves. ❤️ 🗓️ Deadline: Wednesday, October 29, 8:00 PM
23