Jay
@JaySahnan
Founding Eng @ ion (YC W24) & figuring it out in SF.
In January, I dropped everything and moved to San Francisco. 6 months later: - O-1 Visa secured - Joined a YC-Backed Startup as a Founding Eng - Building with @samraaj and @sbarnholdt You can just do things.
Shutting down a startup is harder than starting one. We spent 3 years building and shipping. Late nights. Big wins. Brutal losses. At the start, you're not fully invested. But overtime, it becomes everything. Your identity. Your mission. Your obsession. So when it ends, it…
5am in Shenzhen. this city fuels me to ship hardware like no other.
Day 10 of learning ML for an hour a day Today was kind of slow, I focused in on some more Linear Regression and applied it to some sample datasets Tomorrow is all about Loss Functions
Day 9 of learning ML for an hour a day Didn't know to sum this up in a tweet, so heres a video about Linear Regression.
AI glasses are probably 1-2 years to hit mass adoption. AR at least 5-10 years away.
This video by @LangChainAI explains the fundamentals of context engineering in just 14 minutes. Context Engineering is effectively the #1 job of engineers building AI agents. do not sleep on it.
Day 9 of learning ML for an hour a day Didn't know to sum this up in a tweet, so heres a video about Linear Regression.
Day 8 of learning ML for an hour a day. Today: Training a perceptron, the OG binary classifier. That just means the output is a simple Yes or No (1 or 0). It learns by: 1. Taking inputs 2. Weighing them 3. Making predictions 4. Updating weights when it gets it wrong. Even…
If you’re a founder in SF and need a place to work - DM me. Last week Co-Working was a vibe, join us for the next one. @samraaj @sbarnholdt
i’m in shenzhen, china 🇨🇳 tag your cracked friends, dinners on me 🍻
Day 8 of learning ML for an hour a day. Today: Training a perceptron, the OG binary classifier. That just means the output is a simple Yes or No (1 or 0). It learns by: 1. Taking inputs 2. Weighing them 3. Making predictions 4. Updating weights when it gets it wrong. Even…
Day 7 of learning ML (1 hour/day) Today I learned about the most basic form of a neural net. The Perceptron, inspired by how neurones work in the brain. It takes inputs, weights them, sums them up, and fires if a threshold is passed.
Day 7 of learning ML (1 hour/day) Today I learned about the most basic form of a neural net. The Perceptron, inspired by how neurones work in the brain. It takes inputs, weights them, sums them up, and fires if a threshold is passed.
Day 6 of learning ML (1 hour/day) Today: Why models fail to generalise. If a model generalises well, it has learned the underlying patterns and not just the examples it was shown. It can perform well on new, unseen data. Underfitting -> The model didn't learn well (too…
Last Friday was my last day at Vercel 🥲 I learned so much working with an absolutely cracked team. What makes Vercel special is definitely the people and I’ll miss working with everyone. Big shoutout to @sonofalli for teaching me how to Xeet and some stuff about startups.…
throughout my 8+ years as a startup founder, i've tried 15+ ideas and Jungle is the only one that has had good traction so far. the journey is hard as hell but in the moments i struggled the most, i always reminded myself that as long as i... 1) keep learning from every mistake…
Founder: Our app isn’t growing. Me: How many hours did you spend on GTM last week? Founder: Like 1. Me: I’m revoking your GitHub access, you’re banned from coding this week.
Introducing Bumblebee - the favorite AI teammate built for MSP. Bumblebee lets you talk to your PSA and manage endpoints & alerts - just by chatting in Microsoft Teams. Reply BEE 🐝 for early access! @hire_bumblebee #msp #itsupport