# Do Things That Don’t Scale, and Howard Marks’ Falling-Knife Rule

*By Recommended Reading from Tech Founders • March 8, 2026*

Two genuine recommendations surfaced today, both centered on discipline. Garry Tan revives Paul Graham’s case for manual user contact before automation, while Harry Stebbings points back to Howard Marks’ investor letters for a durable rule in market downturns.

## Most compelling recommendation: product learning before automation

This stands out because Garry Tan does more than restate a classic startup essay. He explains why its core lesson becomes *more* important when AI makes premature scaling easy [^1].

### *Do Things That Don’t Scale* — Paul Graham

- **Content type:** Essay
- **Author/creator:** Paul Graham
- **Link/URL:** Not included in the source material
- **Recommended by:** Garry Tan
- **Recommendation context:** [Garry Tan on X](https://x.com/garrytan/status/2030286543684874590)
- **Key takeaway:** Tan says the essay still matters because its real point is *contact* with users: manual work reveals edge cases, weird behavior, failure modes, hidden dependencies, and the actual jobs to be done before you automate [^1].
- **Why it matters:** In Tan’s framing, good abstractions are discovered by doing the manual loop until the real bottleneck emerges, not designed upfront from a hunch [^1].

> “If you automate before you have that contact, you just scale your misunderstanding faster.” [^1]

## Another useful recommendation: Howard Marks’ market-discipline heuristic

### Howard Marks’ investor letters — Howard Marks

- **Content type:** Investor letters / memos
- **Author/creator:** Howard Marks
- **Link/URL:** Not included in the source material
- **Recommended by:** Harry Stebbings
- **Recommendation context:** [20VC episode](https://www.youtube.com/watch?v=rPaWW0IDurI)
- **Key takeaway:** Stebbings highlights one Marks principle in particular: never try to “catch a falling knife” when assessing buying opportunities during market declines [^2].
- **Why it matters:** It is a compact rule for staying disciplined in downturns, and Stebbings presents Marks’ letters as a recurring source of that judgment [^2].

> “I was always a big fan of Howard Marks and his investor letters and one of his big things is never try and catch a falling knife.” [^2]


[![Mitchell Green: Why 50% of VCs Should Not Exist & Why China will Win the AI War](https://img.youtube.com/vi/rPaWW0IDurI/hqdefault.jpg)](https://youtube.com/watch?v=rPaWW0IDurI&t=298)
*Mitchell Green: Why 50% of VCs Should Not Exist & Why China will Win the AI War (4:58)*


## Pattern across today’s recommendations

Both picks warn against premature action. Tan’s recommendation is about not scaling before direct contact shows what users actually need; Stebbings’ recommendation is about not rushing into a falling market too early [^1][^2].

---

### Sources

[^1]: [𝕏 post by @garrytan](https://x.com/garrytan/status/2030286543684874590)
[^2]: [Mitchell Green: Why 50% of VCs Should Not Exist & Why China will Win the AI War](https://www.youtube.com/watch?v=rPaWW0IDurI)