Paweł Kulak

Thank God.😎👏🏻💨

This is the biggest architectural rethink in conversational AI since Whisper. The 200ms micro-turn design solves the fundamental problem that no one wanted to talk about: every “real-time” voice AI on the market is actually turn-based with VAD bolted on top. That harness is the ceiling. You can’t get to true simultaneous speech, visual proactivity, or time-awareness through scaffolding — the model has to perceive time natively. TML-Interaction-Small (276B MoE, 12B active) scoring 77.8 on FD-bench v1.5 vs GPT-realtime-2.0 minimal at 46.8 isn’t incremental. That’s the gap between “voice interface” and “actual collaboration.” The split architecture is what makes this scalable — interaction model stays present at low latency, background model handles heavy reasoning. Both share context. That’s how you get responsiveness AND intelligence without picking one. The Bitter Lesson framing in the paper is the right call. Hand-crafted dialog management systems will be outpaced by general capability. Anyone still building voice stacks around VAD components is now on borrowed time. Mira’s team just redrew the floor for real-time AI.

@b162543 Interesting.

Worth to check.

X_Quant MetaEnsemble v56 — final OOS audit. If you're at Jane Street / HRT / Jump / Citadel / JPM, this thread is for you. Methodology, not alpha.

@b162543 The outlook of your data looks astonishing!

Poland will see nuclear power entering its energy mix in less than five years as small reactors gain traction in the European country, predicted billionaire Michal Solowow. bloomberg.com/news/articles/…

The developer experience I’ve had with Claude Code over the past months is the reason I stopped using 4 other tools. Subagents, rewind, 1M context, MCP — these aren’t features, they’re a complete rethinking of how you ship code with an LLM in the loop. The fact that the team now has a dedicated channel for changelogs and deep dives tells you where this is going. Day 1 follow.

Hope everyone enjoys their last year of meaningful work!

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.

This has been said a thousand times before, but allow me to add my own voice: the era of humans writing code is over. Disturbing for those of us who identify as SWEs, but no less true. That's not to say SWEs don't have work to do, but writing syntax directly is not it.

I still remember this moment Zelenskyy had with a WW2 veteran in 2024 — it makes me emotional every time. Stand on the right side of history 🇺🇦

@Zsuirad_Nezrok @Propterminalapp Zrobione

Adam Strzembosz. Odszedł Świadek Historii. Człowiek - sumienie naszych czasów. Wieczne odpoczywanie racz mu dać Panie!

@MartaWitecka1 Pani Marto, 7 lat trenowałem pływanie jako junior. Ilość kalorii z tamtego okresu jest już nie do powtórzenia. Powodzenia dla syna!

Piątek z życia pływaka: 5:20 pobudka 5:50 śniadanie 6:30-8:00 basen 8:00-14:35 szkoła 15:00-17:30 basen 18:30-21:00 siłownia,motoryka Gdzieś pomiędzy: jedzenie, picie, jedzenie, białko, jedzenie, picie. 21:15 prysznic, masaż, spanie (21:40) Ja dowożę, odbieram, masuję, gotuję🤯

Do takiego spotkania doszło wczoraj w rezydencji ambasadora w Waszyngtonie, przy okazji uroczystości z okazji Święta Wojska Polskiego. Pan po lewej ma 102 lata, walczył w Powstaniu Warszawskim. Pan po prawej ma 100 lat, walczył pod Monte Cassino. Mieli o czym rozmawiać przy pierogach i polskiej kiełbasie. Wzruszenie? Echhhh. Miło było na nich patrzeć.

It was only ever Arsenal. A boyhood Gooner, our new number 10 – welcome home, Ebere ❤️

To spotkanie na Alasce było lepsze.

@damian_murawiec @OOPRinfo @RuchMF @PRolnikow @OficjalneZero Zmiany klimatu nie istnieją

Sytuacja powodziowa w powiecie elbląskim @OOPRinfo @RuchMF @PRolnikow @OficjalneZero

@TrajektoriaAI Nawet w Chinach rozumieją kapitalizm 🙏🏼🙂‍↕️

Trwają testy wydolnościowe robotów, które miałyby zastąpić pracowników w tym zakładzie mięsnym pod Częstochową. Niestety, na tym etapie nie są jeszcze przystosowane do pracy powyżej 16 godzin dziennie.