One founder built Base44 alone, sleeping in shifts to keep the servers running, and sold it for $80M cash four months later, before hiring anyone. That's a signal for bootstrapped founders in the AI era.
When the One Big Beautiful Bill killed EV and battery incentives, US battery startups lost their demand floor. The Department of Energy just threw them a $500 million line: grants across Coreshell, Lilac Solutions, and Nth Cycle to rebuild domestic supply. The pitch is national security, not climate. Soldiers, drones, and fighter jets need U.S.-sourced batteries, and Washington now treats that as inescapable.
A new Midas Project investigation walks 18 of Andreessen Horowitz's most notorious startups and finds a pattern: bot farms faking human social presence, AI companions linked to abuse, a deepfake platform hosting sexualized content of children, and gambling-fintech built to exploit legal loopholes. Meanwhile the firm is spending tens of millions to shape AI policy, including a $100 million super PAC and alumni in government roles. It wants to write the rulebook it's already breaking.
The DOJ is quietly probing Andreessen Horowitz for holding board seats at competing AI companies, a conflict most VCs assume is routine and rarely enforced. On TechCrunch's Equity podcast, insiders puzzle over why this, of everything, draws a year-long federal investigation. Ben Horowitz sits on Databricks' board; partner Martin Casado on Fivetran's. The real fear: smaller firms get made examples after.
Maor Shlomo built Base44 alone, waking every few hours to mind the servers, and cleared $1.5M in subscriptions within a month of launch. Four months in, no employees, zero VC, Wix bought the vibe-coding startup for $80M cash, with the founder staying on to run it. The deal came together over three dinners at Wix CEO Avishai Abrahami's home. A striking proof that bootstrapped singles can outpace funded teams in the AI era.
Outer Biosciences has quietly spent years perfecting the ability to keep living human skin alive outside the body for over a month. Only now is it going public about training an AI on that living tissue to screen skincare compounds, a test bed far more realistic than conventional cold assays. It is a rare AI-plus-biology gamble that stayed hidden for years.
Jen Abel (JJellyfish, State Affairs) breaks the standard five-stage CRM pipeline, calling it a forecasting tool, not a sales process, and maps the real cycle at roughly 15 steps. The memorable play: a "pincer model" that lands exec and one-level-down meetings simultaneously, plus a two-to-three-sentence pitch built around giving the buyer "alpha" before any demo. Walk-through runs from first cold outreach to procurement and the final signature.
Two-plus months after launch, Anthropic's most capable model pulls in only about 11% of companies' spending on the lab's tools, per Ramp data across 70,000 businesses. Corporate buyers are trading frontier capability for cheaper rivals like DeepSeek that are "good enough." It is a sobering signal for an IPO the market could value at $2 trillion or more.
Hacking ability is the same whether the hat is white or black; only incentives differ. In agentic cybersecurity, offense carries natural advantages, which means defenders must match attackers' capabilities or lose. That dynamic, argues Stratechery, is exactly why AI incumbents, crippled by Washington's restrictions on their own frontier models, are the losers, and startups building offense-ready defenses win.
You’re caught up.
That was all 8 stories in the 24 August issue.