subj
YC buys into the team, more than the idea!
This is one of the most overlooked questions in the YC application:
“How long have the founders known one another and how did you meet?”
Don’t fall into the trap of just saying “6 months” and moving on to the idea, “the more important part”.
Here’s what I noticed during the YC batch: almost every company I knew either completely pivoted or made big adjustments to their idea.
But if the idea is the criterion for acceptance and it changes so often… aren’t these companies just toast?
No. Because a lot of times, YC isn’t buying into the idea at all.
They’re buying into the team. The FOUNDERS.
They want to make sure that you are an exceptional team. That you won’t give up. That you can stick together when you’re stressed and crammed in a room for 16+ hours a day for years (and maybe even living together).
When put like that, can you think of even one person you would want to do that with? If so, maybe it’s time to start a startup lol.
Founder conflict is one of the top causes of startup failure. YC is screening for that.
Some startups pivot 5+ times before landing on their breakout idea. Successful companies will change ideas and directions in weeks. What they won’t change is their founding team.
That’s why this isn’t a throwaway question. It’s your chance to prove you’re the kind of team that will still be standing when the idea inevitably changes. Don’t waste it.
“How long have the founders known one another and how did you meet?”
Don’t fall into the trap of just saying “6 months” and moving on to the idea, “the more important part”.
Here’s what I noticed during the YC batch: almost every company I knew either completely pivoted or made big adjustments to their idea.
But if the idea is the criterion for acceptance and it changes so often… aren’t these companies just toast?
No. Because a lot of times, YC isn’t buying into the idea at all.
They’re buying into the team. The FOUNDERS.
They want to make sure that you are an exceptional team. That you won’t give up. That you can stick together when you’re stressed and crammed in a room for 16+ hours a day for years (and maybe even living together).
When put like that, can you think of even one person you would want to do that with? If so, maybe it’s time to start a startup lol.
Founder conflict is one of the top causes of startup failure. YC is screening for that.
Some startups pivot 5+ times before landing on their breakout idea. Successful companies will change ideas and directions in weeks. What they won’t change is their founding team.
That’s why this isn’t a throwaway question. It’s your chance to prove you’re the kind of team that will still be standing when the idea inevitably changes. Don’t waste it.
subj
Michigan lands #1 RB Savion Hiter!
Boom! Next couple years looking good!
subj
PSA: they’ve been at UofM games for decades. don't worry!
Who are these guys on the roof of the Big House?
That's AAPD, UMPD, and MSP swat. Yes theres absolutely a rifle behind that wall. It's common for large events
That's AAPD, UMPD, and MSP swat. Yes theres absolutely a rifle behind that wall. It's common for large events
subj
the next FoundHERS event
Meet, connect, and be inspired! We’re excited to announce the next FoundHERS event, a series designed to celebrate women’s achievements and amplify diverse voices in entrepreneurship at U-M. Join us for a Fireside Chat with Priya Gogoi (Co-Founder, Celsee, Inc.) and dynamic duo Madeline Eiken & Charlie Childs, Ph.D. (Co-Founders, Intero Biosystems)!
Join us on Tuesday, October 7 from 4:30–6:30pm in Duderstadt 3358. All are welcome! RSVP now: https://myumi.ch/VVdxe
Join us on Tuesday, October 7 from 4:30–6:30pm in Duderstadt 3358. All are welcome! RSVP now: https://myumi.ch/VVdxe
subj
Amazon owes me $427 in refunds for AI books.
But these 9? Worth every penny:
1️⃣ AI Engineering by Chip
Chip built ML at Netflix and NVIDIA. It shows.
She covers the entire stack:
data pipelines → deployment → monitoring → what breaks at 3am.
The chapter on model versioning alone saved my ass twice.
Most books pretend versioning doesn't exist.
2️⃣ Designing Machine Learning Systems (Also Chip)
Her first book. Less AI, more systems thinking.
Covers data drift, retraining, and why your model works Monday but fails Friday.
Reality check: Your model will break. This book shows you how to catch it before users do.
3️⃣ Build a Large Language Model from Scratch
Sebastian doesn't mess around.
No API calls. Just PyTorch and pain. You build a transformer from nothing.
After chapter 3, I finally understood attention. Not the math. The actual implementation.
The truth: 95% won't finish this book. The 5% who do become actual AI engineers.
4️⃣ LLM Engineer's Handbook
Paul and Maxime skip the BS and show production patterns.
RAG that works. Fine-tuning that doesn't overfit. Deployment that doesn't bankrupt you.
Best part: Real code that runs. Not pseudocode garbage.
5️⃣ Building LLMs for Production
Louis-François and Louie bridge the notebook-to-production gap that kills careers.
Scaling, monitoring, cost optimization - the unsexy stuff that matters.
The deployment chapter alone is worth the price. Most books end at training.
Wake-up call: Your Colab notebook means nothing if it can't handle real traffic.
6️⃣ Hands-On Large Language Models
Jay Alammar + Maarten.
Modern tools only. Hugging Face, LangChain, the stack that actually ships.
No reinventing wheels. Just using the best tools correctly.
Why read it: Because you're not building transformers from scratch at work.
7️⃣ Prompt Engineering for LLMs
John & Albert shows why "just be specific" is amateur hour.
Few-shot, chain-of-thought, patterns that survive production. Real strategies, not Twitter tricks.
Cut my error rates by 40%. Turns out there's actual engineering in prompt engineering.
The kicker: Works across all models. Not just OpenAI.
8️⃣ Building Agentic AI Systems
Anjanava and Wrick skip the agent hype and show what actually works.
Multi-step reasoning, tool use, memory systems. Real agent architecture.
Not another chatbot tutorial. Actual autonomous systems.
Reality: Most "AI agents" are just while loops. This shows the real thing.
9️⃣ The AI Engineering Bible
Thomas Caldwell covers everything.
Architecture → infrastructure → team management.
The business side nobody teaches. How to make technical decisions that don't get you fired.
Reads like a CTO's playbook. Because that's what it is.
1️⃣ AI Engineering by Chip
Chip built ML at Netflix and NVIDIA. It shows.
She covers the entire stack:
data pipelines → deployment → monitoring → what breaks at 3am.
The chapter on model versioning alone saved my ass twice.
Most books pretend versioning doesn't exist.
2️⃣ Designing Machine Learning Systems (Also Chip)
Her first book. Less AI, more systems thinking.
Covers data drift, retraining, and why your model works Monday but fails Friday.
Reality check: Your model will break. This book shows you how to catch it before users do.
3️⃣ Build a Large Language Model from Scratch
Sebastian doesn't mess around.
No API calls. Just PyTorch and pain. You build a transformer from nothing.
After chapter 3, I finally understood attention. Not the math. The actual implementation.
The truth: 95% won't finish this book. The 5% who do become actual AI engineers.
4️⃣ LLM Engineer's Handbook
Paul and Maxime skip the BS and show production patterns.
RAG that works. Fine-tuning that doesn't overfit. Deployment that doesn't bankrupt you.
Best part: Real code that runs. Not pseudocode garbage.
5️⃣ Building LLMs for Production
Louis-François and Louie bridge the notebook-to-production gap that kills careers.
Scaling, monitoring, cost optimization - the unsexy stuff that matters.
The deployment chapter alone is worth the price. Most books end at training.
Wake-up call: Your Colab notebook means nothing if it can't handle real traffic.
6️⃣ Hands-On Large Language Models
Jay Alammar + Maarten.
Modern tools only. Hugging Face, LangChain, the stack that actually ships.
No reinventing wheels. Just using the best tools correctly.
Why read it: Because you're not building transformers from scratch at work.
7️⃣ Prompt Engineering for LLMs
John & Albert shows why "just be specific" is amateur hour.
Few-shot, chain-of-thought, patterns that survive production. Real strategies, not Twitter tricks.
Cut my error rates by 40%. Turns out there's actual engineering in prompt engineering.
The kicker: Works across all models. Not just OpenAI.
8️⃣ Building Agentic AI Systems
Anjanava and Wrick skip the agent hype and show what actually works.
Multi-step reasoning, tool use, memory systems. Real agent architecture.
Not another chatbot tutorial. Actual autonomous systems.
Reality: Most "AI agents" are just while loops. This shows the real thing.
9️⃣ The AI Engineering Bible
Thomas Caldwell covers everything.
Architecture → infrastructure → team management.
The business side nobody teaches. How to make technical decisions that don't get you fired.
Reads like a CTO's playbook. Because that's what it is.
express, not stress