🕶️ Self-Taught AI Just Got Real

Ten bullets of updates

1. 🧠 Startups need more than grit. A new guide breaks down the five support systems every founder should build early, from peer mentors to mental health tools and financial advisors.

2. 🪟 Broken window, big breakthrough. After a vandal smashed his storefront, this founder turned the chaos into a viral moment and a major boost in sales.

3.  🏢 This founder made company culture public. By sharing internal values and norms online, they attracted top talent, filtered out bad fits early, and sparked unexpected investor interest.

4. 🧑‍💻 Ex-Shopify founder backs “one-person unicorns”. Birk Jernström’s new startup helps solo developers build scalable SaaS businesses, betting the next big thing might come from a team of one.

5. 📄 Spotting patent-worthy ideas: A new guide helps founders identify innovations worth protecting, with tips on vetting novelty, avoiding common traps, and knowing when to file.

6. 🔥 What makes brands last? After 20 years studying brand success and failure, this expert shares five traits that set enduring companies apart—hint: it’s not just good marketing.

7. 💸 Learning to sell built a $400M startup. This founder credits mastering sales—not product—for scaling to nine figures, proving that storytelling often beats specs in early-stage growth.

8. 🇨🇴 Nubank adds bill pay in Colombia. The fintech giant now lets Colombian users pay utility bills through its app, deepening its footprint in Latin America’s digital banking race.

9. ❤️ Tinder launches double-date feature. The app now lets users team up with friends to match and mingle as a group, aiming to make dating less awkward and more social.

10. 🤖 AI isn’t just taking jobs—it’s making weird new ones. From prompt engineers to chatbot stylists, a new wave of roles is emerging to wrangle, train, and humanize AI systems.

✦

🔎 Congress Eyes Palantir’s ‘Mega-Database’

Ten Democrats led by Sen. Ron Wyden and Rep. Alexandria Ocasio-Cortez just fired off a June 17 letter demanding Palantir reveal every Trump-era contract ($113 million and counting) and explain whether its software is helping the IRS weave a searchable “mega-database” of Americans’ personal data. The lawmakers even likened the project to IBM punch-cards in WWII, warning that a single trove of bank records and medical claims is a “surveillance nightmare” waiting to be abused.

Palantir shot back on X that the Times report fueling the claims is “wildly inaccurate,” adding that it “does not build surveillance technology” and isn’t constructing any central file on citizens.

Why it matters: More than half of Palantir’s $2.9 billion in 2024 revenue came from government work. If Capitol Hill decides the company crossed a privacy line, every data-heavy startup chasing federal dollars suddenly has to wonder where that line actually is and how fast it can shift in an election year.

✦

Startup Budgeting (And What Most Founders Get Wrong)

✦

✦

These companies just raised money

1. 🛡️ Defense startup Mach Industries confirms $100M raise. Just two years old, the AI-powered weapons maker locked funding from Khosla and Bedrock, signaling growing VC appetite for military tech.

2. 🦴 Sword Health raises $40M, delays IPO. The $4B-valued digital therapy startup is pushing its public debut to 2028 or later, citing market uncertainty and a focus on profitability.

3. 📡 Ex-SpaceX founder raises $10M in 36 hours. Andrena, building a faster, cheaper internet network, closed its seed round in record time—aiming to disrupt telecom with tech born in rocket labs.

4. 👗 Alta raises $11M for “Clueless”-style fashion tech. Backed by a16z and celebrity angels, Alta is building a digital closet that recommends outfits and lets users “try on” looks virtually.

5. 📈 Grifin lands $11M to simplify investing. The app auto-buys stock in companies where users shop, turning daily spending into personalized portfolios and aiming to ease newbies into the market.

✦

✦

✦

🧠 AI That Learns (Like, Actually Learns)

Modern LLMs can write poetry, pass exams, and spit out code—but they forget almost everything the moment the conversation ends. A new MIT project called SEAL (Self-Adapting Language Models) might change that. The technique teaches AI to train itself by generating its own notes, reviewing them, and tweaking its internal wiring; kind of like a very nerdy human pulling an all-nighter.

The researchers tested SEAL on open models like Llama and Qwen, and found that the AIs could keep improving well beyond their original training. Instead of just reasoning through tasks in the moment, the models absorbed new context and used it to perform better over time, like a student actually learning instead of just memorizing answers.

It’s not perfect: SEAL is compute-heavy and can trigger “catastrophic forgetting,” where new data overwrites old knowledge (like when you forget your ATM PIN after learning a new Wi-Fi password). But it’s a serious step toward a future where models don’t just respond, they evolve. And maybe, one day, they'll even sleep.

✦

✦

Startup Events and Deadlines

1. Founder Institute Core Program l Deadline: Jun 22 l USA

2. Slidebean Webinar: Don't use YouTube for Startup Marketing l Jun 24 l Online

3. Crash Course in Financial Modeling l Jun 26 l Online

4. Entrepreneur First Bangalore - Fall l Deadline: Jun 29 l Bangalore

5. Slidebean Webinar: Financial Modeling Bootcamp l Jul 1 l Online

6. Entrepreneur First London - Summer | Deadline: July 6 | UK

7. Startupbootcamp | Energy & Sustainability  | Deadline: Aug 10 | South Africa

✦