
Engineer at @Microsoft working on Microsoft Entra. Coder, blogger and sailor ⛵
Brno, Czech Republic
Joined October 2011
- Tweets310
- Following361
- Followers78
- Likes615
Dávid Kaya retweeted
AI interfaces nowadays are still pretty much turn-based.
In my PhD thesis, I thought about what interaction would look like if we could go beyond taking turns.
One example is intra-turn interaction:
Dávid Kaya retweeted
HydraFusion now has an updated model mix pool, reasoning capabilities, and improved orchestration workflows!
... and we heard a lot of asks for a HydraFusion option to throw at your most ambitious work, and introduced a new Max tier to accommodate those tasks!
We appreciate all your continued feedback :)
github.blog/changelog/2026-1…
Dávid Kaya retweeted
Any pursuit of superintelligence has to be grounded in the core principle that if the AI we build is not helping humanity and under human control, it's not worth pursuing.
We also need to accelerate and spread the benefits of AI, such that they are diffused broadly across countries, communities, and companies. This requires a frontier ecosystem in which both closed and open-source models can thrive.
And for firms, it’s imperative that they retain full control over their unique and tacit knowledge. Every organization should be able to build its own continuous learning loop/hill climbing machine, without becoming dependent on any one model provider, and have the ability to embed its own knowledge into models and weights they control.
So, in this context, we welcome the research, focus, and deliberate pacing needed to get alignment right as the design goal. We also welcome ideas like "embedded evaluators" and the broader efforts to develop the mechanisms to make this more than just talk.
The key is that this cannot be controlled by a handful of entities, but must have broad representation across the ecosystem, countries, and fields, including academia.
This is the approach we are taking: broad access and choice at every layer of the AI stack; enterprise control of learning loops and models; and the “Code of Conduct” that underlies our own first party MAI models that we’ll publish tomorrow for public consultation.
4.9 percentage points higher verified task quality. 67% lower estimated cost.
Project HydraFusion delivered those results against Claude Opus 5 on Terminal-Bench 2.1 in controlled offline evaluations.
HydraFusion orchestrates the models and workflow for each coding task. You select it like any other model. ⬇️
github.blog/ai-and-ml/github…
Dávid Kaya retweeted
Project HydraFusion in GitHub Copilot achieves frontier-level quality at up to 67% lower cost!
It achieves this via runtime orchestration that creates a plan, chooses models from multiple providers to draft, critiques and revises, or cascades to more powerful models to complete your task.
github.blog/ai-and-ml/github…
Dávid Kaya retweeted
Excited to see early customers already using Astra on Azure! azure.microsoft.com/en-us/bl…
Dávid Kaya retweeted
We just shipped a new protocol that underlies the VS code agents app and in the future the github app and more
microsoft.github.io/agent-ho…
Dávid Kaya retweeted
Lots of bad information out there --let's clear it up.
• Recent workforce changes were made to restructure the XBOX business because it is not healthy. They were not made to replace employees with foreign workers.
• The H-1B figures being referenced are Microsoft-wide visa renewals and new hire applications. They are not specific to XBOX and represent a small percentage of Microsoft's overall workforce. And the majority of roles impacted were not American roles.
• XBOX is the largest employer of American workers in the gaming industry and the largest American gaming company.
• And Asha is an American born, raised, and educated CEO, from Wisconsin.
Dávid Kaya retweeted
It's time to explore Bohemia in a brand new way!
The official Kingdom Come: Deliverance Board Game, created by @czechgames, is on its way.
Forge your own path through Medieval Bohemia, develop your skills, complete quests, and create your own story, but this time with friends around a table.
Available for Pre-Order now on the KCD Store.
#KCD2 #BoardGame
Dávid Kaya retweeted
I integrated this into Copilot and ran some evals this afternoon. Results were neutral at best and in most cases resulted in MORE token consumption. These results are consistent with our prior evaluation of RTK.
Compression ends up removing information that the agent needs, which triggers a re-read, ultimately resulting in larger costs and latency.
Beware of things that sounds too good to be true.
UP TO 95% TOKEN REDUCTION WITH ZERO CODE CHANGES
A Netflix engineer just open-sourced Headroom, and it’s one of the smartest ways I’ve seen to cut LLM costs.
It wraps Cursor or Claude in a local proxy to compress your payload before it hits the LLM:
→ Intelligently shrinks logs, JSON, and code
→ Perfectly preserves logic accuracy
→ Keeps 100% of your data local
→ Stops Opus-tier models from wasting tokens on boilerplate
It already crossed 35K stars, which says a lot.
100% free and open-source.
repo in 🧵↓
Dávid Kaya retweeted
People should not be judging AI-augmented coding by “1 shots.”
If someone told you that their model did a “one shot of Minecraft” and they’re impressed by that, you need to consider how much semantic heavy lifting the word “Minecraft” is doing in that prompt.
Ask them to one shot Minecraft without using the word Minecraft.
It’s not trivial to one shot something unique, because programming is the art of making the ambiguous incredibly specific through sculpting. AI sculpting is less about vibes and more about finding the specificity you want and keeping the system stable through changes. Good SDLC practices still matter, historical context still matters, and knowing how things work matters, shout out to @Grady_Booch.
It’s a cool party trick to one shot Mario Brothers or space invaders, but then you’ll end up with the most mid version of both. Literally mid. You’ll get the statistical fat part of the Bell curve version of these mythical games. You’re telling the model to close its eyes and draw the face of these games from memory.
As high-level programming cedes way to the prose compiler, making your goals and specs well understood to the ambiguity loop and showing good judgment is going to matter more than ever. Consider all of your words and make sure that certain words aren’t carrying all the semantic load, hidden or otherwise.
Today we introduced Maia 200, our next-generation AI accelerator now powering production workloads in Microsoft’s U.S. datacenter regions delivering faster, more efficient AI inference at scale. Learn more: msft.it/6012Q1Hx0
64
Dávid Kaya retweeted
Today we announced our new Fairwater datacenter in Atlanta, connected with our first Fairwater site in Wisconsin and our broader Azure footprint to create the world’s first AI superfactory.
Fairwater exemplifies our vision for a fungible fleet: infra that can serve any workload, anywhere, on fit-for-purpose accelerators and network paths, with maximum performance and efficiency.
AI workloads have evolved beyond large-scale pre-training. Today, they encompass fine-tuning, reinforcement learning (RL), synthetic data generation, evaluation pipelines, and more. Fairwater is built to support this full lifecycle:
Max density: Fairwater’s two-story design and liquid cooling system lets us place racks in three dimensions and pack them with GPUs as densely as possible, minimizing cable runs and improving latency and effective bandwidth.
Fleet: Each Fairwater DC can integrate hundreds of thousands of the latest NVIDIA GPUs into a single coherent cluster. This provides flexible infra that can support the full spectrum of workloads, and ensure no GPU is left unnecessarily idle.
And that’s on top of the more than 100,000 GB300s coming online this quarter alone for inference across the rest of our fleet. For us, it’s all about turning every gigawatt into the maximum number of useful tokens. Not every GW is created equal!
Planet-scale: Every Fairwater DC will connect through our continent-spanning AI WAN to prior generations of AI supercomputers, forming a truly fungible pool of compute. This enables developers to scale beyond the capacity of a single site and dynamically land workloads on the right infra for their needs.
Together, these innovations let us bring together different generations of silicon and AI systems across DCs and geos into a single elastic system that scales seamlessly across training and inference workloads
And this elastic AI capacity is all available alongside all the other cloud services (compute, storage, databases, app services) that AI agents and workloads need.
This is what we mean when we talk about building a fungible fleet – a single, unified platform that pushes the limits of performance per watt and per dollar.
Read more: blogs.microsoft.com/blog/202…
Dávid Kaya retweeted
During a recent government visit to Kuttenberg, Mayor Lukáš Seifert surprised the Czech Prime Minister @P_Fiala with a Kingdom Come: Deliverance 2 Collector’s Edition🤯
We are incredibly proud to see our game presented at such a level and grateful for the support and recognition it receives within Czech society.
(We were told the Prime Minister’s team burst into jealous applause. Can’t blame them! 😜)
Dávid Kaya retweeted
Agent Mode for Visual Studio! Give it a natural language prompt for the task at hand and GitHub Copilot will create a multi-step plan and iterate until it gets the job done!
youtu.be/7_duh0HoT9o
Dávid Kaya retweeted
C# is programming language of the year 2023 per the TIOBE Index! #dotnet tiobe.com/tiobe-index/
























