Airbound just banked $37M on a bet that drones should weigh less than what they carry. And if you think AI agents will let you sleep more, the founders pulling all-nighters because of them have news for you.
Airbound just banked a $37M Series A from Greenoaks, DoorDash, and Lachy Groom on a counterintuitive wager: flight is only cheap when your aircraft weighs less than the payload. Its 3.3-pound TRT drone already hauls 2.2 pounds of diagnostic samples for Narayana Health, cutting a multi-hour road trip to seven minutes. The fly-through goal is cost parity with trucking across an Andhra Pradesh metro network.
Cheap AI tooling flips the build-vs-buy math, but Goldsmith warns the real cost was never the build. His startup's hand-rolled auth worked until the maintenance hit: constant security patches, enterprise features the vendor would ship free, and hours in customer security reviews. Cheap to build, expensive to own forever. The question isn't what it costs to make, but who patches it in year three.
Replit's CEO now expects more than half his company to be salespeople by year-end, and SaaStr says the pattern holds across the AI cohort. The catch is timing: PLG companies once added sales around $20M-$50M ARR, now it's more like $100M-$250M. Gamma hit $100M with no sales team and its CEO calls that a mistake. Anthropic added 140+ salespeople in 18 months and now posts more sales than research roles.
There is no "best" GTM motion in 2026, just the right one for your numbers. Inbound leads adoption at 23%, while product-led growth dominates below $1M ARR and $5K ACV, and account-based wins for products over $25K. No motion predicts faster growth. This field guide walks founders through the levers that do: channels, AI GTM infrastructure, outbound signals, and working pricing models.
Founders are pulling all-nighters not despite their AI agents but because of them. One co-founder's bots ship features at 2 a.m., so he's gone to bed at 6 a.m.; another, retired for good, un-retired and now clocks 7:30 a.m. to 2 a.m. daily. The more agents do, the more human work their speed creates. The irony: automation marketed as leverage is delivering agents-and-bone hours.
Solo founder Ryan Carson ran 15 concurrent Devin agents for a month at a $20k bill across engineering, customer success, and investor updates. His radical take: manage them with folders and a paper list, not a dashboard, and stop chasing more output. The interview breaks down the "Watchdog" playbook that replaced a CS team, when he still reaches for Codex, and why a single real customer meeting changed his product direction.
Paul Graham argues that writing isn't just a way to convey ideas, it's a way to have them. Read deeply and you get the raw material; reading well is the only route to writing well, and writing is the only route to thinking clearly about hard, ill-defined problems. People who only want information may find faster shortcuts. People who want original ideas cannot afford to. (464 words)
A founder handed free access to 10 prospects who'd all described his exact problem, and none of them ever logged in. That's the tell: price was never the barrier, desire was. This plain-English primer explains why, arguing product-market fit is a retention curve, not a feeling, and that a cohort bleeding to zero usually means you're acquiring the wrong users, not that the product is mispriced.
You’re caught up.
That was all 8 stories in the 25 August issue.