Phat T. Pham

@emtalas I’m in

I have a bit of a crazy idea: A 48-hour distribution hacker house in Muskoka Small group of founders, we go deep on: - content systems - personal brand - inbound + hiring through distribution You leave with a repeatable growth engine + early signs it’s working Does this sound interesting to anyone? 👀

@hasantoxr All AI notetakes be like: 👀

🚨 BREAKING: Microsoft just open-sourced a frontier Voice AI that handles 60-minute audio in a single pass. You drop in a recording. It identifies every speaker, timestamps every word, and outputs a full structured transcript with who said what and when. It also does real-time TTS with just 300ms first-audio latency and supports 50+ languages. 100% Open Source.

@jasonlk What does that mean when you date a girl in sales

Never hire anyone in sales that doesn't care about money You'll see

@IlirAliu_ Solving the teleoperator problem is an important prerequisite to solving the model problem.

Teleoperation today relies on WebRTC. That works for Zoom. Not for robotics. Is that statement provocative? WebRTC works for video calls. But not really for robots. I’m teleoperating from SF → London using Adamo’s stack. Zero buffering. No lag smoothing… just real-time control. Because WebRTC apps add buffers to handle: • packet loss • network jitter That’s fine when you’re talking. Not so much when you’re controlling a robot. Even small delays kill precision. So instead of adapting WebRTC… Adamo built a custom protocol + dedicated global infrastructure. Without buffering or artificial delay. A stable, real-time connection across the Atlantic. This is what people underestimate about robotics: Models are just one piece of the puzzle. But looking at the entire stack, there is so much more. And Sometimes you can’t reuse what exists. You have to build it from the ground up. Thank you, Sam Considine for letting me try, it nice seeing you during GTC last week! 👋 ——- Weekly robotics and AI insights. Subscribe free: 22astronauts.com

Notion is becoming Granola and Granola is becoming Notion

@vivoplt Where’s your distribution cost

- Claude = coding. ($20/mo) - Supabase = backend. (Free) - Vercel = deploying. (Free) - Namecheap = domain. ($12/yr) - Stripe = payments. (2.9%/transaction) - GitHub = version control. (Free) - Resend = emails. (Free) - Clerk = auth. (Free) - Cloudflare = DNS. (Free) - PostHog = analytics. (Free) - Sentry = error tracking. (Free) - Upstash = Redis. (Free) - Pinecone = vector DB. (Free) Total monthly cost to run a startup: ~$20 There has never been a cheaper time to build.

@cormachayden_ With AI, everyone can be at that intersection as learning is easy.

the most dangerous person in any room: - the engineer who understands sales - the marketer who can code - the finance person who can tell stories - the founder who makes content - the salesperson who can write none of them are the best at any one thing they just own an intersection nobody else has the ability to articulate yourself makes you dangerous in every role

@niccruzpatane There's 8B people on the planet so the humanoid robots will replace humans several times over.

@mehedih_ Split the transcripts by topic of conversations. So you have a thread of topics that is time based.

ok so I really liked how Granola worked so I had Composer 3.6 build my own Granola for me - it auto detects meetings and transcribes all ur meetings and u can even ask questions during the meeting, check it out here: granola.ai what features should i add next?

Just in! Unitree files for a ~$610M IPO after delivering 5,500+ humanoid robots, ~1.7B RMB in revenue, and gross margins above 60%, making it the only company in the category operating at scale with profitability. This IPO sets the first public market benchmark for humanoid robotics and begins to define how the sector will be valued, funded, and expanded going forward.

Looping on LLM to make agents is just a bandaid approach to make it streaming.

@ScottWu46 @dabit3 Can it try to reproduce the bugs before attempting it to fix it? Otherwise it’s hallucinating the bug and solving for a non existent bug that results in a bloated codebase

This week 70% of all Devins were started by humans (webapp, slack, linear) and 30% were started automatically (API, and now scheduled + managed Devins) In a few months that probably flips to 30/70 the other way and within a year it'll be 10/90. What does it look like to run a truly agent-native dev team? -Handing whole project specs to a top-level agent that manages its subagents to do the groundwork -Agents that kick off automatically on Sentry/Datadog alerts and as first-line incident-response -Agents continuously running integration test / QA workflows and auto-investigating UX changes or performance issues And of course, all of the "eng scaffolding" needed to make this possible - comprehensive unit testing, good documentation, and a reproducible dev environment.

@NotionHQ The only app that can compete with Granola

🎙️ Update: AI Meeting Notes now run in the background. Switch apps, lock your screen…doesn't matter. Your notes will keep going. On iOS and Android.

@bpizzacalla @mattshumer_ AI becomes the managers

@mattshumer_ the wild part is we spent two years debating whether AI takes jobs, and doordash just quietly made "employed by an AI agent" an actual job category. the relationship flipped and most people haven't even noticed yet.

DoorDash is laying the groundwork for a crazy move here. Agents will be able to 'hire' humans to do tasks for them in the real world. And this will collect insane amounts of training data for robotics. Kind of genius, kind of terrifying.

@mattshumer_ Truly passive businesses here we come!!

@andyfang Can my AI agent assign task and pay a human to do them?

Introducing Dasher Tasks Dashers can now get paid to do general tasks. We think this will be huge for building the frontier of physical intelligence. Look forward to seeing where this goes!

@andyfang GG taskrabbit

@tommycollison Do you guys have the LLM to detect whenever there’s retries and grab the the correct code to prompt the LLM to produce a dataset sample for fine tuning?

Composer 2 is now available in Cursor. It's frontier-level at coding, priced at $0.50/M input and $2.50/M output. There's also a faster variant with the same intelligence at $1.50/M input and $7.50/M output. Deets below!

What’s really powerful? Is that whenever an LM apologizes or get some things wrong they should create a data asset sample for training or a fine tuning later on.

@RohithKolluri__ @perplexity_ai How do we know for certain it doesn't write anything to the databases?

I haven't written a line of SQL in months With @perplexity_ai's Computer, it feels like I have a full team of data scientists working 24/7 Analysis that used to take me hours is done in minutes now It's connected to my data warehouse, runs dozens of queries in parallel, and doesn't stop until I have all the info I need Feels superhuman