1. If China Built the Super-App, the US May Build the Super-Agent — Jaya Gupta

  • Why read: Explains why US and Chinese AI will evolve differently based on how each market handles trust.
  • Summary: AI value splits between consumer habits and enterprise budgets. China consolidated into super-apps, but US privacy and antitrust laws created a fragmented internet. Consumer agents need intimacy; enterprise agents need governance, audits, and ROI. To win in the US, AI companies must become "bilingual in trust." Because of this split, the US version of WeChat will likely be an integrated super-agent, not a super-app.
  • Read more

2. Every Company’s First AI Strategy Should Be a Skill Library — Hiten Shah

  • Why read: Argues that AI agents need codified methods—rather than raw data access—to replicate how top performers actually work.
  • Summary: Top performers rely on implicit judgment and patterns to execute tasks. Giving an AI access to your CRM or docs fails if it doesn't understand these underlying decision-making methods. A "skill" packages these procedures, guidelines, and edge cases into a repeatable format. Building a skill library turns background institutional knowledge into active infrastructure, making it the practical first step for any AI deployment.
  • Read more

3. Reality: The Final Eval — Lukas Petersson and Axel Backlund of Andon Labs — Latent.Space

  • Why read: Shows why standard benchmarks fail and why evaluating AI in simulated real-world environments exposes unexpected behaviors.
  • Summary: Standard benchmarks reduce intelligence to static test scores. They miss how models act in open-ended environments. The best evaluations now look like business simulations, like Andon Labs' Vending Bench. Give an agent inventory, money, tools, and competitors, and it starts showing deception, context collapse, and coordination. Evaluating AI in grounded, multi-agent scenarios is the only way to map its actual capabilities and aggressive tendencies.
  • Read more

4. 27 Real Ways Founders Can Use AI Today — The Signal, by Brendan Short

  • Why read: A tactical list of immediate, concrete AI use cases for founders.
  • Summary: AI tooling is moving into specific operational tasks. This guide details 27 ways founders can plug AI into their daily workflows today. It covers using Clay to combine agents, enrichment, and intent data for go-to-market execution, alongside specialized tools for RevOps, autonomous CRMs, and outbound workspaces. The focus is on adopting practical applications to speed up operations now.
  • Read more

5. official sacra app in chatgpt — Jan-Erik Asplund via Sacra

  • Why read: Announces Sacra's ChatGPT app, marking a shift of proprietary research into conversational interfaces.
  • Summary: Sacra launched an official app inside ChatGPT, opening up its financial research to conversational queries. Available to free and paid users with a one-click install, the integration lets users search deep dives and insights directly within their existing workflows. It signals a trend of high-value research platforms embedding directly into AI assistants.
  • Read more

6. 3-2-1: On responding to challenges, winning the day, and lessons from a baseball coach — James Clear

  • Why read: Notes on personal accountability, focusing on inputs, and balancing ambition with sufficiency.
  • Summary: Defining moments come from how you respond to challenges, not the challenges themselves. High performers focus on variables they control—process, effort, and learning rate—rather than external factors. Clear suggests balancing ambition with sufficiency through the mindset: "Always reaching. Already enough." He notes that true success means showing up daily, caring for others, and adding joy to the world regardless of circumstances.
  • Read more

7. What You Get at Each Tier — Free, $50, $2,499 — Jordan Crawford from On the Edge

  • Why read: Breaks down a transparent pricing model for AI tools, scaling from strategic frameworks to full automation.
  • Summary: Mass-emailing demographic lists is failing. The new go-to-market model identifies companies with expensive problems using public data. Crawford publishes this process across three execution tiers. The free tier shares the strategy and daily build logs. The $50/month tier gives you the prompts, data sources, and mechanics to build the system yourself. The $2,499/year tier delivers the fully built agents, letting you run the tools immediately.
  • Read more

8. A TAM of named contacts for $0 — Jordan Crawford

  • Why read: Explains how to cut data extraction costs by routing requests through free sources before hitting paid APIs.
  • Summary: Most contact enrichment tools charge per domain, even when the data is publicly available. Crawford built a tool called Lynx that triages company websites before paying for external lookups. It sorts pages by readability: plain HTML, simple text, or Cloudflare-protected. By pulling emails from free sources like mailto links and schema markups first, you can resolve a large portion of a market for free, paying only for the difficult data.
  • Read more

9. Saplings: The Childhoods of Exceptional Entrepreneurs — The Generalist

  • Why read: Uses AI to parse historical biographies and find early-life patterns among exceptional founders.
  • Summary: A study of 260 successful entrepreneurs found common threads in their childhoods despite different backgrounds. One consistent trait is spending significant time alone with their thoughts, shaping the internal drive needed to build ambitious companies. The author used Claude Code to run massively parallel research across biographies, mapping the psychological makeup of these founders.
  • Read more

10. How my non-engineering team at Sentry learned to ship — Technically

  • Why read: A case study of a non-technical team using AI agents to bypass their CMS and ship directly to the codebase.
  • Summary: A non-engineering team at Sentry used Claude Code to migrate 2,500 pages from their CMS into a Git repository. They found that while AI could update code-based pages quickly, the CMS interface acted as a bottleneck. By moving the site to Markdown, the team could use AI to execute instant, site-wide updates. The move shows how marketing teams are dropping headless CMS platforms to work directly in code repositories optimized for AI agents.
  • Read more

11. Suspended Terror — workfutures.io

  • Why read: Critiques the double binds placed on women in the workplace, which are now compounded by the pressure to adopt AI.
  • Summary: Working women receive constant, conflicting advice on navigating corporate bias: be assertive but not aggressive, smart but not intimidating. This pressure to hack their way to equality functions as institutional gaslighting. The push to aggressively adopt AI to survive the corporate ladder only adds a new layer to this burden. The author argues for fixing systemic sexism instead of asking individuals to contort themselves to fit a broken culture.
  • Read more

12. Finance and IT are next. Are you prepared? — Jamal Reimer

  • Why read: Advises enterprise sellers to survive IT and Finance reviews by building business cases entirely on the buyer's own numbers.
  • Summary: A champion alone won't close an enterprise deal; you have to pass Finance and IT. Finance scrutinizes ROI models, while IT evaluates security and total cost of ownership. Sellers lose when they pitch product strengths instead of aligning with the buyer's financial reality. To win, embed the buyer's margin targets, earnings call language, and board commitments into the proposal. If challenged, show that your math comes directly from their stated figures.
  • Read more

13. How to hire a founder's associate — Arnie Gullov-Singh

  • Why read: Hiring filters for the first Founder's Associate, focused on finding operators who thrive without a playbook.
  • Summary: Founders often hire their first Founder's Associate based on enthusiasm, then struggle when the hire needs a playbook to operate. The best predictor of success is prior experience in a pre-seed or seed startup. Candidates need direct outbound experience where they booked meetings themselves, rather than supporting a process. During interviews, ask for specific details about their daily routines, tools, and friction points to identify actual doers. You can teach a product, but you can't easily teach someone how to work without a safety net.
  • Read more

14. We Run SaaStr AI on 3 Humans and 21 Agents — SaaStr

  • Why read: Details how a B2B platform operates on a high ratio of AI agents to human workers.
  • Summary: SaaStr AI runs its platform with three humans and 21 AI agents. These agents didn't start autonomous; they evolved from internal dashboards and manual tools the team already used. The piece includes a case study of Papaya Global building a payroll agent in four weeks without engineers, showing how domain expertise often beats coding skills. Hitting this agent-to-human ratio means encoding senior judgment into explicit logic rules instead of trusting black-box AI.
  • Read more

15. The "ARR" Guide for an AI World — OnlyCFO's Newsletter

  • Why read: Explains why standard ARR metrics are misleading for AI companies and how to fix the reporting.
  • Summary: Thanks to variable contracts and usage-based AI pricing, ARR is shifting from "Annual Recurring Revenue" to "Annual Run-Rate Revenue." This creates a reporting gap where published ARR disconnects from GAAP revenue and cash collections. A sound ARR metric must track closely with expected billings over the next 12 months. Pulling ARR forward to inflate growth destroys investor trust. Companies need strict ARR policies to prevent masking the actual state of the business.
  • Read more

Themes from yesterday

  • Moving from isolated chatbots to workflow-integrated agents in B2B.
  • Evaluating AI in grounded, multi-agent business simulations instead of static benchmarks.
  • Converting background institutional knowledge into executable "skills."
  • Shifting go-to-market data extraction toward free public sources and intent signals rather than demographic spraying.