YC’s AI Binge, Zombie Unicorns & the Self-Building Startup

Startups that hit $1 billion valuations during the zero‑interest era are now too weak to IPO or be bought, leaving employees and investors in limbo. Their stalled growth traps capital and inflates risk across venture portfolios, signalling a broader correction in tech financing.

YC has doubled down on AI, inflating batch sizes to nearly 200 startups and backing many low‑moat B2B AI projects. Critics argue the accelerator’s due‑diligence is slipping, risking founder reputations and investor confidence. The shift could reshape how early‑stage capital screens for sustainable advantage.

Polsia, an "AI Slop" startup, says it hit $10 million in annualized revenue with zero employees, using AI agents to run every business function. YC‑backed Thomas follows a similar model, where an AI founder launches and manages companies for a cut. The trend hints at a new class of firms that sell company‑building as a product, reshaping how startups are created and funded.

OpenClaw’s pull‑request volume exploded from two a week to 3,400, while merge rates fell from 48 % to under 10 %. Most new PRs are AI‑generated spam submitted in seconds, eroding repo trust. The crisis pushes projects toward sender‑reputation tools such as Mitchell Hashimoto’s Vouch.

PayPal deployed Salesforce’s Agentforce AI SDR to cold‑call roughly 8,000 monthly leads that reps never touched. The AI ran a ten‑touch cadence, handing qualified meetings to reps, and conversion rates rose about 50% versus human‑only effort. It shows AI agents can unlock revenue in buried pipeline without reshaping data first.

AI inference resellers lose margin if they charge cost‑plus, but switching to outcome‑based pricing and cutting compute costs can restore 30‑plus‑point margins. Tom Tunguz shows how model routing, caching, and distillation decouple revenue from raw token prices, letting startups sell work, not compute.

Cursor grew to $100 M ARR in under two years by turning VS Code into a drop‑in AI‑powered fork, so developers kept every extension and shortcut they’d built up. By removing any need to switch tools, it captured a massive share of VS Code’s user base. The lesson: erase switching costs and the customers come with themselves.

AI agent startups that promise all‑in‑one automation are seeing sign‑ups evaporate when their products churn out junk outputs. The post argues that narrow, thoroughly tested features beat grandiose claims, or else massive churn will sink valuations and future funding.

Moats aren’t permanent walls; they’re a habit of relentless, low‑margin execution that competitors hate. Amazon’s thin‑margin model shows a moat survives only while the company endures uncomfortable repetition. The insight: sustainable advantage is a disciplined, continuous process, not a static asset.

Robin Guo reflects on his first six months as a founder after leaving a16z, highlighting the autonomy and conviction that come with running a biotech startup. He contrasts the experimental, B2B‑driven shift at his former venture fund with the hands‑on decision‑making and learning curve of building a company from scratch.

Open‑weight models such as DeepSeek V4 are priced up to 50 times lower than Anthropic’s and OpenAI’s frontier offerings, dramatically undercutting the API‑only business model. This cost shock forces proprietary providers to slash prices or invent scarcity tactics, reshaping the economics of AI deployment.

Adobe announced it will postpone a $500 million ARR price hike for Creative Cloud, signaling that customers are finally pushing back after years of relentless increases. The move reveals a tension between raising rates and acquiring free‑user volume, hinting that B2B pricing power may be eroding faster than expected.

SignalFire data shows engineering hiring fell only 11% while overall tech hiring dropped 25% in 2025, and engineers accounted for 55% of new hires at major tech firms, up from 46% in 2019. The finding contradicts AI‑driven job‑loss narratives and suggests engineers remain in high demand across both big firms and startups.