Our internal data shows Claude is accelerating AI development—a possible path to recursive self-improvement, or AI autonomously building a more capable successor.
It’s happening faster than we thought, and the implications deserve greater attention. https://t.co/OVVPJO7VQx
Almost no speakers from .@sama perhaps from all the drama with .@elonmusk , .@satyanadella had enough. .@mustafasuleyman had quietly been building the models. But no custom silicon to compete and provide cheaper alternatives to azure users, so .@sundarpichai is still winning.
$MSFT
Microsoft held Build 2026 and the message was clear they’re done being OpenAI’s distribution partner
New reasoning model. New hardware called Project Solara. Agents that run autonomously across your entire workflow. Their own dev PC with Nvidia’s RTX Spark chip
They’re not adding AI to software anymore. Software is being replaced by agents
MSFT ran this same playbook before. They won the OS era. They won the cloud era. Now they’re setting up to own the agent era
A prototype that is nowhere close to being in actual products just to say that you are also in the league of ai labs and Google but you actually aren't even close?
One last run-through before Build tomorrow when I can finally share what we’ve been up to in the lab. Livestream starts at 9:30 am PT, and you can register or tune in here: https://t.co/HUPJqU0QMF
“Ex-Microsoft exec says the company blew it with Al, as it did with mobile”
"Not even 3% of paying Copilot users use it even when it's pre-deployed right in their faces”
The Microsoft 3% problem. See Word and Excel features.
Advances in LLMs created a super interesting problem space and recommendation models and systems seem less exciting. Are there any successful stories with crazy ROI in the intersection of LLMs and Recommendation models? Can your entire ranking system boil down to a single LLM?
Super app super flop. First they need to get their shit right with what they already have. Copilot and copilot 365 and copilot in outlook. Terrible ux compared to chatgpt claude
Huge if true. But it isn't. We all know how it goes with papa .@satyanadella . He's not looking to innovate and stay at the frontier. He's trying to use enterprise lockins and capitalize on sales and marketing strategies.
Microsoft is launching homegrown AI models at Build next week, positioned as cheaper alternatives to OpenAI and Anthropic.
Buried in the reporting: relying on Anthropic's Claude forced Microsoft to raise GitHub Copilot prices and cap how much developers could actually use it. They get OpenAI's models free through 2032 but are spending billions to not need them when that expires.
Suleyman's team hasn't topped a single leaderboard in two year, the explanation being that Microsoft's OpenAI deal restricted him from training frontier models until April. Convenient timing for a launch.
Via The Information
Anyways, really excited for Microsoft Build next week. Ill attend in person.
We've raised $65 billion in Series H funding at a $965 billion post-money valuation, led by @AltimeterCap, Dragoneer, @Greenoaks, and @sequoia.
This investment will help us advance our research and expand our capacity to meet growing demand for Claude.
We shouldn't even care about leaders' words from successful firms. Forget about failed firms like Microsoft AI. The AI .@mustafasuleyman and team do is build applications on top of .@sama 's models. And yes, they have their image model in the top ten after a decade of work.
Easy. Just stop listening to these CEOs. Filter out the noise. That's all they have to add. Noise. Let the work of their companies and numbers speak for themselves.
Most of that was for help with translation from logic you have in mind or worked out in pesudo code, to match target APIs and language syntax. There was some thinking outsourcing as well but it wasn't as easy as it is today. Rarely did what you want to do exactly match a SO q&a.
Thankfully this is in summer when Seattle shines so much better than the Bay Area. .@satyanadella You better show off your foundational models or custom silicon, or a QPU. Even an open model that is gaining adoption will be nice.
Microsoft's @satyanadella kicks off Microsoft Build 2026 on June 2 at 930am PT. We'll be sharing what's new across AI, agents, and more. Save the Date!
https://t.co/yGOUHrDVFW
Today we’re releasing DeepSWE, a new standard for agentic coding benchmarks.
On public leaderboards, top models often look relatively close in capability. DeepSWE shows where they actually diverge, reflecting the realistic experience of developers in their day-to-day work.
Good God, I haven't seen a statement as inaccurate as this. CA is the worst in taxes utilities everything period. If your paycheck is majorly going to Uber grocery, you have a different problem than most who don't go