# Top 10 in Tech - What to know for Week ending July 25, 2025

Published: 24 July 2025
Canonical: https://www.top10in.tech/posts/week-ending-july-25-2025

## 1. SaaS METRIC OF THE WEEK

Here is a highly bookmarkable [Guide to SaaS Metrics](https://equals.com/guides/saas-metrics/) from [equals.com](https://equals.com) that covers all the greatest hits and more (ARPA, LTV:CAC, Burn Multiples, etc).

Link: https://equals.com/guides/saas-metrics/

## 2. STARTUP PLAYBOOK

Check out the [Emerging Startup Playbook](https://www.growthunhinged.com/p/the-emerging-startup-playbook) that lays out eight key emerging strategies taken from high-growth SaaS leaders. Emphasizing "speed over perfection," it suggests streamlining decision-making and embracing usage-based pricing for sustainable growth. Key takeaways? Optimize product stickiness before scaling sales and pivot based on data to stay competitive. This is a great read for early-stage founders figuring out their growth tactics (well, more like experiments).

Link: https://www.growthunhinged.com/p/the-emerging-startup-playbook

## 3. AI BUDGET

Here is a quick, fun one. How much should you spend on AI? Tomasz Tunguz says early-stage startups are [allocating 20–50% of their R&D budget to AI](https://tomtunguz.com/ai-rd-percent/). Nope - not a typo. AI isn't just a feature, it's a core investment now. Just pretty sure you can't claim those R&D costs back in any tax breaks.

Link: https://tomtunguz.com/ai-rd-percent/

## 4. SOFTWARE 3.0

Stuff is moving so fast these days, it's hard to keep up. [This post](https://substack.com/home/post/p-168696477) (taken in inspiration from [these slides](https://docs.google.com/presentation/d/1sZqMAoIJDxz79cbC5ap5v9jknYH4Aa9cFFaWL8Rids4/edit?slide=id.g33d19f2dc57_0_424#slide=id.g33d19f2dc57_0_424)) makes the case that we've moved beyond Software 2.0 (ML/neural networks) to Software 3.0: AI-native and conducted in your native language, with API-connected agents reasoning over dynamic data. Tools like LangChain and LlamaIndex aren't the future—they're gonna be the foundation.

Link: https://substack.com/home/post/p-168696477

## 5. SPEND

Brex has launched something called the [Brex Benchmark](https://www.brex.com/journal/brex-benchmark-may-2025), and so far, the May 2025 Benchmark is a gem: it's real spend data from millions of anonymized transactions. Despite economic headwinds, ad spend is up — a signal of renewed growth bets. But spend on consultants and real estate is falling. SaaS and infra? Holding steady with Claude being the numero uno AI app.

Link: https://www.brex.com/journal/brex-benchmark-may-2025

## 6. JOBS

AI may be a primary culprit for a bunch of the current economic payoffs. A new [CNBC analysis](https://www.cnbc.com/2025/07/20/in-job-losses-ais-role-may-be-bigger-than-companies-say.html) shows AI is playing a much bigger role in job losses than companies admit. The reshaping of work is happening faster than we're told.

Link: https://www.cnbc.com/2025/07/20/in-job-losses-ais-role-may-be-bigger-than-companies-say.html

## 7. ACCOUNTING AI

This one is a bit of a mind-bender from Penrose and explores if LLMs can do accounting and [how LLMs can transform ledgers into language](https://accounting.penrose.com/), enabling a new term "explainable accounting". It's a weird and dense page, but super interesting for any of you finance ops peeps. BTW - [Claude is already on it](https://www.anthropic.com/news/claude-for-financial-services#fromHistory)!

Link: https://accounting.penrose.com/

## 8. PERSONAL CO-PILOT

This is def another AI-heavy newsletter—but [here's how to build an AI co-pilot for your day-to-day](https://www.lennysnewsletter.com/p/build-your-personal-ai-copilot). From task routing and meeting prep to knowledge surfacing and auto-recaps, Lenny's Newsletter shares a stack of tools to make AI actually useful (and sticky.....and on-brand) in your workflow. [ChatGPT now has a next-level co-pilot](https://openai.com/index/introducing-chatgpt-agent/), an agent that can actually perform tasks on your behalf.

Link: https://www.lennysnewsletter.com/p/build-your-personal-ai-copilot

## 9. OUTAGES

We all have them - especially if we host in Public Clouds - for example, [last month GCP service control tripped](https://amriksh.medium.com/lessons-in-cloud-resilience-unpacking-june-12-2025-gcp-outage-6db1adc575d3), creating outages across GCP and Google Workspace, triggering Cloudflare issues, which in turn triggered AWS and Azure outages - it was wild. [This piece digs into why reliability is so hard at scale](https://newsletter.pragmaticengineer.com/p/why-reliability-is-hard-at-scale)—even for top-tier teams. From incident fatigue to architectural tradeoffs, it's a sharp look at how engineering orgs really manage risk (or don't until hindsight). A must-read if you don't know how resilient or redundant your stack is.

Link: https://amriksh.medium.com/lessons-in-cloud-resilience-unpacking-june-12-2025-gcp-outage-6db1adc575d3

## 10. CASE STUDY: ONBOARDING

AI onboarding isn't just UX—it's growth. [This teardown](https://www.fishmanafnewsletter.com/p/how-ai-products-drive-adoption-in-onboarding-through-template-activation-loop) shows how ChatGPT, Claude, and others use template-activation loops to turn cold signups into active users.

Link: https://www.fishmanafnewsletter.com/p/how-ai-products-drive-adoption-in-onboarding-through-template-activation-loop

## POD OF THE WEEK

Want to know what Software 3.0 (number 4 above) may look like in real life - [this dude has been figuring all this out since only January](https://www.lennysnewsletter.com/p/successfully-coding-with-ai-in-large-enterprises) and presents an extensive (and IMO quite compelling) overview of the new AI-based management paradigm. Makes me feel really far behind.

Link: https://www.lennysnewsletter.com/p/successfully-coding-with-ai-in-large-enterprises