Top 10 in Tech - What to know for Week ending August 14, 2026

SaaS METRIC OF THE WEEK

Churn.: See #2 for more on Churn\*\*.\*\* According to CatchJS, though, we're all calculating churn rates wrong. If you love Statistics, the article is well worth reading, and it even gives some Python code to perform the more complicated probability-based equation they recommend. You can then check out this tool (as a handy Google Sheet) from Newfund as a way to analyze the strength of revenue streams for any B2B startup. A complimentary article outlining the methodology behind the tool is here (and you should read it first).

CHURN

See #1 above\*\*;\*\* it's the ultimate leaky bucket. BVP's guide on tackling customer churn explains how to identify root causes and implement strategies to reduce attrition (that can be terminal). Analyzing churn data, improving customer onboarding, and enhancing product value to retain users are all in there. The guide also provides actionable steps for creating a comprehensive churn action plan to plug those leaks.

INVESTMENT

Hey founders - check yourself (at least your metrics) - Investors have an experienced nose for this stuff, and generally, legitimately good metrics only mean something in pairs. NRR next to GRR tells you if retention is real or propped up by your upsells. Burn multiple at 1.1x - nice, that's top quartile. Now look at Rule of 40 next to it, down from 38% to 22% in four quarters? That's not efficiency, that's a company that stopped spending and stopped growing at the same time.

ONBOARDING (STAFF)

Only 12% of employees reckon their company onboards them well, so check out these 7 steps for onboarding a new team member (and they start two weeks before day one). Real handy if your onboarding currently means a laptop, slap on the back, and a Slack invite.

VENTURE

Carta's VC Fund Performance Q1 2026 is out. $3.9B raised across 86 new funds, the strongest start to a year since 2022, and TVPI (Total Value to Paid-In) is up for nearly every fund vintage. All on paper though. The median 2017 fund has returned 0.31x to its LPs after nine years, and 2018 sits at 0.15x.

AGENTS

Computer-using agents went from 42% on OSWorld a year ago to 85% today (which is past the 72% humans score). Before you get outraged and/or excited - 85% means 15 of every 100 tasks failed, and a business process only finishes if every step does. $6-8 an hour is what a16z reckons it costs to run a computer-using agent, against $10 for offshore people and $30-45 for a US back-office worker.

CPU

Turns out the agent boom isn't only a GPU story - AWS engineers were told in May to cut CPU waste and now reportedly wait days for an EC2 instance that used to take hours - agentic workloads are making CPUs scarce. Apparently, Intel, AMD, and others say customers will take whatever they can get.

VISIBILITY

Your sales are flat so you post more, add a platform, say yes to another podcast......etc. Sylvia Huang calls it "the visibility treadmill" - more output, same revenue, tireder founder. The real question underneath: have enough of the right people actually seen this yet? Until they have, you can't tell a reach problem from a product problem.

TRUST

Report time and ignore the Plaid bias evident in this - rest of it is good. 44% of Americans would let an AI agent execute trades, it goes up to 54% for Gen Z and Millennials. But on the other hand, 80% of Americans think companies should reimburse them when AI makes a mistake with their money, and 78% say AI should be held to the same accountability standards as a human advisor. Can you have your AI cake and eat it?

CASE STUDY

M&A -: Airtable just sold to Bending Spoons for $1.3B, and they last raised at $11.7B in 2021! CJ Gustafson pulled 10 lessons out of it, and they're all worth your time, but here's the one that stings: back-of-the-envelope math from 20vc says nobody who invested after March 2018 made money (that includes employees).