Preetham Nadig

I build technology that pays for itself across data, AI & agentic systems. Writing on the agentic shift at Digital Reflections.

[Preetham Nadig @PreethamNadig]

Pricing is the decision I revisited most as a founder — and got wrong the longest.

Data and AI platforms are structurally harder to price than the software before them. Three reasons. 🧵

[Preetham Nadig @PreethamNadig]

False comfort:

"We're mission critical" — so is the database.
"Deep domain expertise" — if it lives in the UI, the call leaves it behind.
"Customers won't switch" — switching used to mean a migration project. https://t.co/C4gW9Ovifq

[Preetham Nadig @PreethamNadig]

The same decision — exposing your product as callable tools — multiplies some companies and hollows out others.

The sorting isn't about execution quality. Four tests tell you which side you're on. 🧵

[Preetham Nadig @PreethamNadig]

3. Is there a network on your side — counterparties, rails, a shared reference set?

4. Who carries the accountability when it goes wrong? That's the one the harness structurally doesn't want.

[Preetham Nadig @PreethamNadig]

6. You get disintermediated from the outcome. The harness holds the goal and reports the result; you supply a step.

Only two of these are really about MCP. The rest was always the shape of the business. https://t.co/TfRL9qEl7P

[Preetham Nadig @PreethamNadig]

When incumbents plan an MCP server it gets scoped as an integration project: which endpoints, what schema, how auth works.

All necessary. All beside the point. Six things dissolve, and the plan accounts for none of them. 🧵

[Preetham Nadig @PreethamNadig]

4. Seats stop matching reality — one agent carries many logins' work.

5. Substitution becomes a config change. Interchangeability is the POINT of a protocol.

[Preetham Nadig @PreethamNadig]

The version I'd defend:

Opening MCP doesn't cause obsolescence. It reveals whether you were ever more than a convenient interface over a commodity capability.

Same decision, opposite outcomes. https://t.co/BrvVGAZGyg

[Preetham Nadig @PreethamNadig]

Almost every software incumbent is shipping an MCP server right now.

Correct decision. Also, for many of them, the moment their product stops being a product and becomes a component in someone else's workflow. 🧵

[Preetham Nadig @PreethamNadig]

"So don't open it?" That fails.

Withholding doesn't keep work in your UI. It routes around you — a competitor who did open, scraping, automation driving your own screens. You lose distribution and stay just as exposed.

[Preetham Nadig @PreethamNadig]

Same sequence running now, faster.

Don't aim to be earliest. Aim to still be standing, somewhere durable, when the dust settles.

Time in market is not a moat. Being first is not a position. https://t.co/5f1zyIMlLh

[Preetham Nadig @PreethamNadig]

I founded a startup into the Big Data wave believing being early was the whole advantage.

It isn't. Early mostly buys you the bill for the market's education. 🧵

[Preetham Nadig @PreethamNadig]

Then the dust settles, and narrowly. Big Data had enormous variety; cloud plus a few stacks was the long play.

Being early doesn't protect you from the shakeout. Consolidation rewards whoever owns a durable layer, not whoever got there first.