
Research @XBOX | PhD @ucsc | prev @Microsoft @Meta | I can help with Software, UX Research, Design, Games Research, AI | All views/reshares are personal
Redmond
Joined December 2016
- Tweets1K
- Following2.6K
- Followers1.1K
- Likes1.2K
I led the XBOX research case study @satyanadella shares here - while the cost and speed gains are incredible, it was awesome to use MD1.1 confidence outputs to know when to triage data to bigger models, identify severity etc
Microsoft-Decision-1.1 is now generally available in Microsoft Foundry and through OpenRouter and Vercel for fast, low-cost decision making.
We've added multimodal support and it's 2.2x faster with improved accuracy. It's powering experiences across Microsoft and used by more than 4,000 customers since public preview.
3
9
597
It's been around 4 months since I joined XBOX Research and I can not believe something I did out of passion and desire to learn more about players made it to the big screen at #MSIgnite
This XBOX use case for Microsoft Decision models is really cool. They switched from using GPT 6 Sol to Microsoft Decision 1 to sort player feedback into themes, and it was 14x faster and 200x less expensive. #MSIgnite
6
148
I see all the love EA Cricket'07 coming in and I am glad, I love the game, it has my childhood - but sad to see the hate @BigAntStudios is getting - they make good games, I play them regularly it's one of my top stress busters and it's good. @RossSymons your craft is awesome!
3
130
While this is getting popular pulling back some of my early papers from UG. It plays around with camera angles to auto-detect cricket deliveries and stitch highlights, and build play-graphs - I often say think this paper gave me admission for a Games PhD dl.acm.org/doi/pdf/10.1145/3…
cricket 07 with an online multiplayer mode. play with your friends
all you need is a web browser
link: cricket07.dhaiwat.com
1
3
202
Sai Maram retweeted
I've been cooking up retro vibes in XBOX LABS. Tell @asha_shar and I your favorite era!
I was lucky to get access, run tests and use Microsoft-Decision-1 for some studies - It is an awesome model for classification, sorting etc!
Introducing Microsoft-Decision-1, our new model for fast decision-making.
It delivers top performance on structured decision tasks, outperforming both LLMs and other decision models in latency and quality.
We’re already testing it across Microsoft for everything from incident response and quality control to scientific discovery.
1
132
Sai Maram retweeted
Great to see Microsoft-Decision-1 as the smartest model from the most attractive quadrant, with the highest non-hallucination rate among decision models. A strong fit for real customer use cases, which is what we trained it for.
Replying to @ArtificialAnlys
Across the four benchmarks in the Artificial Analysis Decision Index v1.0, Perplexity’s Decider v1.1 27B has the lowest Cost per 1M Tasks at $17. It sits alongside Microsoft-Decision-1 ($31) and Jev 1.13 ($42) on the Capability vs. Cost Pareto frontier. Celeris-1 Decision has the fastest Time per Task at 0.06 seconds, 2x faster than Jev 1.13 (0.13 seconds), which is 11x faster than Claude Haiku 5.5 (Low) (1.4 seconds)
I like the idea behind @typesafeai Jev, spent some time building an UX Research affinity mapping tool. Can not wait to have an explain/reasoning parameter in the response would add so much value for UX. But gosh the speed is insane! @CompleteSkeptic and team awesome work!
2
129
Sai Maram retweeted
Your physical game library is more than a collection of discs. It holds years of memories and stories worth preserving. Starting next week, XBOX Insiders can unlock a digital entitlement for supported physical games. We’ll start with thousands of titles and add more as we can.
Sai Maram retweeted
hello there the jacobian conjecture is false thanx to my close friend akhil for asking about it and my other close friend fable for working during the world cup final
((1+xy)^3 z + y^2 (1+xy) (4+3xy), y + 3 x (1+xy)^2 z + 3 x y^2 (4+3xy), 2 x - 3 x^2 y - x^3 z): \C^3\to \C^3, has jacobian determinant -2, and sends (0, 0, -1/4), (1, -3/2, 13/2), and (-1, 3/2, 13/2) to (-1/4, 0, 0)
Can someone please explain how all these node-edge based canvases for design are worth millions and billions of dollars? Like what is the moat? I built this over a weekend; and I cannot help but imagine, fable 5 making this in a single go at this point. xitter.gq/ADesiHCI/status/197599…
For too long, our best ideas have been scattered across tools and platforms. Every idea deserves space to grow, experiment, and materialize. Your imagination is limitless—your tools shouldn't be.
Introducing Ooha (ఊహ) — the Telugu word for imagination. A canvas that connects, expands, and supports your limitless imagination.
✨ Connect. Expand. Create.
Try it now: ooha-creator.github.io/
(You need your own API keys)
#AI #OpenAI #GeminiAI #CreativeTools #Innovation @OpenAIDevs @perplexity_ai @google
1
194

















