
our first experiments into how to move past tools that feel like factories: studia.co

Partner @StratMindsVC. Prev @WhatsApp design. Consciousness-UX — aboutblank.pub. Bicycle for the soul.

our first experiments into how to move past tools that feel like factories: studia.co


What a week in the San Francisco Bay Area! I met my partners in crime @bragiel & @danbragiel, portfolio founders @sonia_piorek @LesyaHendrix @yfilipch @AntonPolishko @arturkiulian Pati Brzozowska, many old and new friends like @tyggy @neffko @cparjaszewski @thisismosalah @promethe @calikasia @KielakKacper @Igor__Markov @mikitamikado & more! It's been too long since I last visited so I'm making the pledge to come to SF *at least* once a year from now on. Off to New York City today on the same mission of (re)connecting the Polish, Ukrainian and CEE diaspora. See you soon!

Introducing Selfdrive. The AI growth team for your iOS app that gets you users while you sleep. It handles Apple Ads, ASO, SEO and UGC creators for you. All of it. superapp.dev/selfdrive

fascinating timeline we're in
3/ A common shortcut is to train on the post-trained model. Unfortunately, this often “fry” the model: it degrades capabilities, destabilizes the model’s preferences, and causes reality drift—the model becomes confused about what’s fake and what’s real.

The paradox is that once a model is able to oneshot something, it becomes worthless. Doesn't matter how cool it is, since everyone will produce exactly the same and nobody wants to consume it.
Opus 5.5 on Max effort - "make a dynamic 15-second motion graphics video that shows what an incredible motion designer you are, like it's your showreel for a résumé. go all out."

Today, we’re introducing a new way to create in @getjamboree: Prompting works great when you know exactly what to ask for. But most people just don't. So we came up with a radical idea: A graphical user interface. Tap, don’t type.


Cosmos has been honored in FastCompany's 2026 Innovation by Design Awards ✨ Design has always been core to Cosmos — from the interface to the systems behind search and discovery. Thank you @FastCompany for the recognition.

What happens as the cost of creating software goes to zero?


We’re releasing speculative decoding in our inference engine Uzu, starting with Qwen3.6-27B. On Apple M5 Max with 128 GB of unified memory, our Mirai-M stack reaches 105 output tokens/sec entirely on-device - 2.9× faster than the fastest MLX speculative-decoding implementation we benchmarked. The result comes from full-stack co-design: DFlash + our Weaver model, tree-based speculative decoding, Mirai quantization, our verification algorithm, and Metal kernels optimized for Apple silicon. Mirai builds the full stack for frontier on-device intelligence. 105 t/s is the evidence, full-stack integration is the moat. But tokens/sec are not actually the metric we ultimately care about. More on that soon. Full benchmarks + methodology: trymirai.com/metrics

Today we are releasing our speculative decoding implementation in our inference engine uzu. Initially for Qwen3.6 27B, with support for Qwen3.8 27B and Muse Glimmer coming soon. On Apple M5-series chips, we outperform MTPLX (MLX + speculative decoding) by almost 2x, and llama.cpp by over 3x at comparable quantization levels, with the strongest gains achieved on mathematical reasoning and coding tasks. Run the model: Mirai-M: trymirai.com/local-models/al… Mirai-L: trymirai.com/local-models/al… Explore the benchmarks: trymirai.com/metrics Learn more about our speculative decoding implementation: trymirai.com/blog/speculativ… Our draft model, quantized checkpoint format, verification algorithm, and GPU kernels are co-designed from the ground up around the latest Apple M5 chips to take maximum advantage of GPU Neural Accelerators. Unlike popular speculative decoding architectures such as model-native MTP, which produce small draft chains of 3-4 tokens at a time, we use extremely aggressive speculative budgets of 16-32 tokens. This enables us to use Neural Accelerator-backed GEMM kernels, achieving maximum utilization of hardware arithmetic throughput.

Same
I’ll be going to the Digital Minds weekend in person in San Francisco!!


we hosted “foraging ep01” a themed app preview who was most locked in? @Behzadhaghgoo @Reshusaur @eunifiedworld @covacut @PradyumnaShome @tyggy @farayofsun @diyasabh @ben_issen @eve_bouff @barofclo
i was locked in bro

AI put coding in your pocket. Now it’s design’s turn. Meet Playful, the designer on your phone. Create when inspiration hits, from a party or on the couch. Have an upcoming party / event? We're creating the next 888 event covers for free. RT + tag @playfuldesignco with your request⭐