Taner Topal

We just launched Flower Hub. A persistent problem in federated learning is that the app layer never really had a proper home. Discovery was weak, sharing was clunky, and trust was mostly implicit. Flower Hub changes that by making collaboration around apps much easier. It gives the community a place to publish Flower Apps, review them, and star them. That makes trust more explicit and portable, so others can run the same apps with much less friction. The result is a simpler way to share trusted training pipelines for collaborative AI. I’m especially interested in what this could unlock for projects like autoresearch, where the ability to exchange and run training pipelines across distributed data feels fundamental. Would love to hear what you think.

I'm leaving Germany | Brutally Honest Review

Proud to share that today, we're launching Flower Hub; a place for publishing, discovering, and running Flower apps across heterogeneous environments 🎉🎊. More than 100 apps are already live on Flower Hub from day one. Personally excited about this launch! 🚀😁

Flower makes it very easy to start with federated learning without worrying about any complex setup. For local simulation, there are basically two commands you need to care about: • one to generate the app — flwr new and • one to run it—flwr run

Flower Labs and Starcloud Reach a Major AI Milestone in Orbit AI infrastructure is expanding beyond Earth, with space emerging as a key part of where AI is headed next. Today, @flwrlabs and @Starcloud_ are sharing an important step forward: the successful execution of an AI workload using Flower on an operational Starcloud satellite, with telemetry retrieved from orbit confirming end-to-end execution. The mission included locally fine-tuning a ViT transformer to classify high-resolution satellite imagery, urban areas, forests, lakes, and more -- directly on the satellite where the data is generated. The result points to a new paradigm for AI infrastructure. Model training and updates no longer need to rely exclusively on centralized cloud data centers; they can take place in space. This is only the beginning. We look forward to building larger, decentralized AI networks in orbit in continued collaboration with Starcloud. Full Blog Post Announcement: flower.ai/blog/2026-02-0… @ycombinator

Similar experience here: I can work in unfamiliar stacks by describing intent and validating outcomes. I can usually judge correctness, but without the LLM I wouldn't know which APIs to reach for or even what to search. Feels like if you know what you want, you can build it in almost any ecosystem like suddenly being fluent in every language.

A few random notes from claude coding quite a bit last few weeks. Coding workflow. Given the latest lift in LLM coding capability, like many others I rapidly went from about 80% manual+autocomplete coding and 20% agents in November to 80% agent coding and 20% edits+touchups in December. i.e. I really am mostly programming in English now, a bit sheepishly telling the LLM what code to write... in words. It hurts the ego a bit but the power to operate over software in large "code actions" is just too net useful, especially once you adapt to it, configure it, learn to use it, and wrap your head around what it can and cannot do. This is easily the biggest change to my basic coding workflow in ~2 decades of programming and it happened over the course of a few weeks. I'd expect something similar to be happening to well into double digit percent of engineers out there, while the awareness of it in the general population feels well into low single digit percent. IDEs/agent swarms/fallability. Both the "no need for IDE anymore" hype and the "agent swarm" hype is imo too much for right now. The models definitely still make mistakes and if you have any code you actually care about I would watch them like a hawk, in a nice large IDE on the side. The mistakes have changed a lot - they are not simple syntax errors anymore, they are subtle conceptual errors that a slightly sloppy, hasty junior dev might do. The most common category is that the models make wrong assumptions on your behalf and just run along with them without checking. They also don't manage their confusion, they don't seek clarifications, they don't surface inconsistencies, they don't present tradeoffs, they don't push back when they should, and they are still a little too sycophantic. Things get better in plan mode, but there is some need for a lightweight inline plan mode. They also really like to overcomplicate code and APIs, they bloat abstractions, they don't clean up dead code after themselves, etc. They will implement an inefficient, bloated, brittle construction over 1000 lines of code and it's up to you to be like "umm couldn't you just do this instead?" and they will be like "of course!" and immediately cut it down to 100 lines. They still sometimes change/remove comments and code they don't like or don't sufficiently understand as side effects, even if it is orthogonal to the task at hand. All of this happens despite a few simple attempts to fix it via instructions in CLAUDE . md. Despite all these issues, it is still a net huge improvement and it's very difficult to imagine going back to manual coding. TLDR everyone has their developing flow, my current is a small few CC sessions on the left in ghostty windows/tabs and an IDE on the right for viewing the code + manual edits. Tenacity. It's so interesting to watch an agent relentlessly work at something. They never get tired, they never get demoralized, they just keep going and trying things where a person would have given up long ago to fight another day. It's a "feel the AGI" moment to watch it struggle with something for a long time just to come out victorious 30 minutes later. You realize that stamina is a core bottleneck to work and that with LLMs in hand it has been dramatically increased. Speedups. It's not clear how to measure the "speedup" of LLM assistance. Certainly I feel net way faster at what I was going to do, but the main effect is that I do a lot more than I was going to do because 1) I can code up all kinds of things that just wouldn't have been worth coding before and 2) I can approach code that I couldn't work on before because of knowledge/skill issue. So certainly it's speedup, but it's possibly a lot more an expansion. Leverage. LLMs are exceptionally good at looping until they meet specific goals and this is where most of the "feel the AGI" magic is to be found. Don't tell it what to do, give it success criteria and watch it go. Get it to write tests first and then pass them. Put it in the loop with a browser MCP. Write the naive algorithm that is very likely correct first, then ask it to optimize it while preserving correctness. Change your approach from imperative to declarative to get the agents looping longer and gain leverage. Fun. I didn't anticipate that with agents programming feels *more* fun because a lot of the fill in the blanks drudgery is removed and what remains is the creative part. I also feel less blocked/stuck (which is not fun) and I experience a lot more courage because there's almost always a way to work hand in hand with it to make some positive progress. I have seen the opposite sentiment from other people too; LLM coding will split up engineers based on those who primarily liked coding and those who primarily liked building. Atrophy. I've already noticed that I am slowly starting to atrophy my ability to write code manually. Generation (writing code) and discrimination (reading code) are different capabilities in the brain. Largely due to all the little mostly syntactic details involved in programming, you can review code just fine even if you struggle to write it. Slopacolypse. I am bracing for 2026 as the year of the slopacolypse across all of github, substack, arxiv, X/instagram, and generally all digital media. We're also going to see a lot more AI hype productivity theater (is that even possible?), on the side of actual, real improvements. Questions. A few of the questions on my mind: - What happens to the "10X engineer" - the ratio of productivity between the mean and the max engineer? It's quite possible that this grows *a lot*. - Armed with LLMs, do generalists increasingly outperform specialists? LLMs are a lot better at fill in the blanks (the micro) than grand strategy (the macro). - What does LLM coding feel like in the future? Is it like playing StarCraft? Playing Factorio? Playing music? - How much of society is bottlenecked by digital knowledge work? TLDR Where does this leave us? LLM agent capabilities (Claude & Codex especially) have crossed some kind of threshold of coherence around December 2025 and caused a phase shift in software engineering and closely related. The intelligence part suddenly feels quite a bit ahead of all the rest of it - integrations (tools, knowledge), the necessity for new organizational workflows, processes, diffusion more generally. 2026 is going to be a high energy year as the industry metabolizes the new capability.

🌼 Flower Labs is hiring a Senior DevOps Engineer to help shape the future of open-source AI. In this role, you’ll design, automate, and operate the infrastructure that powers Flower’s open-source and enterprise platforms, the leading framework for federated and decentralized AI, trusted by organizations like @mozilla, @OwkinScience, and @jpmorgan If you’re passionate about building automation, optimizing systems, and contributing to the world’s most popular open-source federated learning framework, this role offers a chance to make real impact in AI infrastructure. 👉 Apply here: buff.ly/qVRLb7i

🚀 Berlin, get ready to federate! We’re thrilled to co-host our next hackathon. Together with @exalsius, Technische Universität Berlin, and the Einstein Center Digital Future (@ECDigitalFuture), we’re launching Cold Start: Distributed AI Hack Berlin. Join us on November 14 & 15 for Europe’s first hackathon focused on federated and decentralized AI. 🎯 Your mission: Use a distributed GPU setup (prepared and provided by exalsius and Flower Labs) + curated datasets to train the best-performing federated model. ⏱️ Nov 14 & 15 📍 Einstein Center Digital Future, Wilhelmstraße 67, 10117 Berlin 🔗 RSVP: buff.ly/2mbqZbD ⚠️ Limited spots -> apply fast! 🍕 Drinks, pizza, lunch & good vibes Join us and shape the future of distributed AI.

🚀 New Paper Alert: FlowerTune: A Cross-Domain Benchmark for Federated Fine-Tuning of LLMs — from @flwrlabs We're thrilled to introduce FlowerTune, the first benchmark that evaluates federated fine-tuning of 26+ LLMs across 4 domains: NLP, Finance, Medical, & Code.

We’re excited to announce that @AMD is sponsoring our upcoming Decentralized AI Hackathon by Flower event. AMD will be providing two Radeon GPU that will be added the existing prize pool. 🎉 Looking forward to seeing so many at @Stanford next Friday for this sold out event. 🏆 Big thanks to AMD for supporting the community and empowering builders, researchers, and students to experiment with decentralized AI methods. 📍Jen-hsun Huang Engineering Center, Stanford University 📅 26 September 2025 🔗 More details: buff.ly/ZPd7a7s

🌼 Flower AI Day is Coming 🌼 The future of AI is federated and @flwrlabs is bringing together leaders, builders, and visionaries to shape it. 🎤 Hear from pioneers from @cohere, @jpmorgan, @Meta, @togethercompute, @GoogleDeepMind, @Stanford and more! 🚀 Join us for a day of cutting-edge talks, hands-on demos, and community connections. 📍 @SHACK15sf, Ferry Building - San Francisco 📅 September 25, 2025 🎟️ Reserve your spot now: luma.com/446t20pc 🔗 Watch the promo video below and see why you can’t miss it 👇 #FederatedLearning #AI #Innovation

The Flower Enterprise Launch Stream is live! x.com/i/broadcasts/1…

🚨 The Flower Enterprise Launch Livestream kicks off today — and we’re going live with Session 1 in just 30 minutes! Built on the trusted open-source Flower framework, Flower Enterprise is designed to accelerate the real-world adoption of federated AI at scale. It delivers the security, scalability, and deployment tools that large organizations need to bring federated systems into production. 🎥 Here’s what’s coming up in Session 1: → Welcome and Introduction to Flower Enterprise  🌼 Daniel Beutel (@daniel_janes) -- Co-founder and CEO, Flower Labs → Special enterprise guests:  🧬 Nikolas Pontikos (@npontikos) -- Co-founder, @Eye2Gene  💡 Matthieu Blottière -- Tech Lead, @OwkinScience → Live Demo of Flower Enterprise  🌼 Chong Shen Ng (@chongshenng) and Javier Fernandez (@jafermarq) -- Flower Labs If you can't make Session 1, join us for one of the other sessions later today! 🏗️ Learn more about Flower Enterprise: buff.ly/jfOELl2 ☎️ Book a demo: buff.ly/3ehPwtx 📺 Livestream on Flower Social: LinkedIn · X · YouTube

😎 Think AI is cool? 🗓️No plans for September 25, 2025? Well, now you do. 🌼Flower AI Day 2025 is taking place in San Fransisco (@SHACK15sf) on September 25. It's the largest federated AI conference - now on the west coast as well. 🎙️Bunch of cool speakers from @Meta, @cohere, @Stanford, and more to be announced ... 🎟️ Get your tickets here: flower.ai/events/flower-… 📰 Read more here: medium.com/p/flower-ai-da… @flwrlabs

It was a pleasure working with the amazing Kinexys team at @jpmorgan led by Sudhir Upadhyay on this ImageNet-like moonshot for financial AI. More than a year of work -- congrats to Sudhir, his team and collaborators @BNYglobal. Glad that @flwrlabs could act as an enabler for project AIKYA.

Join us in #SanFrancisco for a mega day full of #decentralized #Ai and #opensource

in San Fransisco? @flwrlabs is going to SF this September! Much to be expected so don't miss this! We'll bring announcements and features to you live, cool tech, and amazing speakers and keynotes. Expect talks from @flwrlabs ' own @daniel_janes & @niclane7 , Joseph Spisak at @Meta , Timothy Chou at @Stanford, and maybe more surprise guests ... 👀 See you there!

@readswithravi Couldn't agree more. People often forget that perfect is the enemy of good enough.

🌸 Just had a 🔥 convo with Daniel Beutel, founder of @flwrlabs — they’re building a new kind of AI: decentralized + federated. Instead of sending your data to the cloud, Flower runs AI on your device: ⚡️ Faster 🔒 Private 📴 Works offline 🧠 More personalized It’s like Apple Intelligence — but open source, dev-friendly, and ready for healthcare, finance, and enterprise. Also, its developer conference is running now and they are live streaming it on its X channel. Just one of the decentralized AI companies on my list, I'll post that in a reply.

Announcing COLLECTIVE-1: the first user-owned foundation model, made possible by @vana and @flwrlabs. Specs: 7B parameters, using private data from 1M+ users on Vana - data that no single company has. Goal: scale to 100M users to create better AI than any centralized AI lab.