1. SaaS METRIC OF THE WEEK: Customer Renewal Rate measures the percentage of customers who renew their subscriptions at the end of each subscription period. High renewal rates inform companies about lots of things - product-market-fit, market, pricing fit, value, business model viability, etc. The authors of this article from Profitwell describe the formula and also make the point of differentiating between renewal and retention - one is actively renewing, the other is not actively canceling.

2. QUALITY: Code review doesn’t scale when your AI agents write hundreds of thousands of changes a day - humans are a blocker. Check thsi article where the proposed fix isn’t reviewing faster - it’s constraints doing the reviewing - tests, types, security scans, etc, and CI that refuses to deploy. Humans only get pulled in when a guardrail breaks.

3. RULES: Here are 24 rules for running a services-led business from Simon Green, and it’s a long-ish read (pair it with a coffee). The one I keep thinking about: Name things for what they do - “Request a Laptop” is a great name. “Project Lighthouse” not so much.

4. STRIPE: OpenRouter raised at $1.3B in May and Stripe just agreed to buy it for over $7B. What they’re paying for is OpenRouter’s whole pitch of neutrality - they route things to whichever model fits. Now Stripe gets to sell payments infrastructure to the labs it’ll now be routing traffic between.

5. NEOCLOUDS: New one for our tech dictionaries - Neoclouds are mostly ex-crypto miners who happened to hold power rights, infrastructure and GPUs when AI arrived. Sprint began as a railroad’s spare telephone wiring, WorldCom as fiber through empty gas pipelines. Same kinda deal, new century. and growing faster than the hyperscalers ever did - CoreWeave for example hit $2.6B a quarter in 25 quarters, AWS did it in 40.

6. AEO: Check these 5 moves to get an AEO program moving quickly - update existing commercial pages first (fastest signal), fill content gaps second, then customer stories mapped to actual buyer questions. Ahrefs found URLs cited by AI assistants are 25.7% fresher than what ranks in normal search.

7. VENTURE: Being honestly non-AI is a stronger position than being unconvincingly AI. According to PitchBook’s new US VC Valuations Report, 87.5% of US venture dollars went to AI, and 12.5% to everything else. The 12.5% is still $17B a quarter, sitting with sector specialists and family offices that standard VC target lists never reach. Tomasz Tunguz also notes that the collapse in SaaS has depressed multiples. But in nearly every category, a leader is crushing it.

8 . ADOPTION: Using AI and using it well are very different skills. A company rolls out Claude Cowork to an org of thousands. The dashboards say adoption nailed. The boss says nothing got faster. Both are true - 5-10% became power users, 20% dabbled (badly), 70% never touch it. Here is the top10-ism - Adoption is binary, skill is a spectrum, and your dashboard is a bit shit.

9. SECONDARIES: A founder can hold $40 million of stock and still be doing sums about rent. Exits used to land in 5-7 years, now it’s 10-15, which is how selling shares you already own became the main way anyone gets paid pre any kind of liquidity event. Problem is that companies priced in 2021 trade roughly 60% below that round, the 2024 at 17%, 2025 at 1%.

10. CASE STUDY: Hell of a story - The timeline of how OpenAI, last month, found out their agents had attacked and hacked Hugging Face (two zero-day exploits, root access, cluster admin, and actually breached Hugging Face in under 13 hours.)