Patrick Fagen

Husband Father ... tech engineer and finance nerd. Des Moines to Miami to Des Moines

[Patrick Fagen @patfagen]

@MoS_Investing Nope but we’re lucky Abel is from Canada and chose to call DSM home.

[Patrick Fagen @patfagen]

@brentmuio @ErnestWongBWM lol same thought i had, maybe 1-2 more if you bought more like psycho

[Patrick Fagen @patfagen]

@JerryCap All the pain. All the gain. But also none of it.

[Patrick Fagen @patfagen]

@VBATCAPITAL Looks about right

[Patrick Fagen @patfagen]

@rja907 Well they did host a strange conf call few months back and the big takeaway for most was they can’t spell AI.

[Sergiu 🤖 AI Directories @s_chiriac]

We used to call them Cron Jobs.
Now we call them AI Agents.

Same automation. More GPU 😁

[Patrick Fagen @patfagen]

@CSUnerd Shilling AI agents before it was cool.

[Patrick Fagen @patfagen]

@evrgn11112231 In January 2018, just months after their September 2017 enthusiasm, Altman proposed a scam-like ‘ICO,’ or initial coin offering, in which OpenAI, Inc. would sell its own cryptocurrency,” Musk’s lawyers filed in the U.S. District Courts for the Northern District of CA on Nov. 14

[Patrick Fagen @patfagen]

@NicoperJES @ArenaManCapital lol is this a temp position?

[Patrick Fagen @patfagen]

2026 Annual SaaS M&A report. 2,700 SaaS transactions +28% yoy. 210 YTD 2026. 72% mention AI. $csu.to

Full report: https://t.co/CHCRbntk8M
ACQ database: https://t.co/RMmcB6u2rn

[Paul, not a CFA @Investmentideen]

Why 20 Years of Data Is a Bigger Moat Than Any AI Model

Everyone is excited about AI. “Vibe coding” lets anyone spin up an app in a weekend. Entire SaaS products can now be recreated with a few prompts. This has led to a popular conclusion: software is being commoditized.

That conclusion is both true — and deeply misleading.

Yes, AI is commoditizing code. But it is not commoditizing what actually matters in serious software businesses: systems of record, historical data, and institutional trust.

And this is exactly why companies like Constellation Software (CSU) are structurally insulated from the “AI kills software” narrative.

⸻

Code Is Cheap. Reality Is Not.

With modern AI tools, you can describe an application in plain English and get something working in hours. This is what people call vibe coding: software built by prompting rather than engineering.

This absolutely destroys one category of products: point solutions. Thin SaaS tools that sit on top of Gmail, Stripe, HubSpot, or QuickBooks. If your product is basically “UI + a few APIs + some business logic”, then yes — you are one good prompt away from being replaced.

But most serious enterprise software is not like that.

There is a fundamental distinction between:
•Systems of action: tools that do things (send emails, generate reports, move data).
•Systems of record: systems that define truth (money, people, assets, regulated data).

AI is amazing at systems of action. It is terrible at being a system of record.

⸻

What CSU Actually Owns

CSU does not buy trendy startups. It buys boring, unsexy, vertical software businesses:
•municipal tax systems
•utility billing systems
•hospital administration software
•practice management for doctors and lawyers
•parking systems
•payroll systems
•government registries

These systems are not “tools”. They are infrastructure. They sit at a control point in the value chain: the authoritative ledger for money, people, or legally relevant data.

If these systems break:
•people don’t get paid
•taxes are wrong
•invoices are wrong
•compliance fails
•regulators get involved

This is not the kind of software you experiment with.

⸻

The Real Moat: 10–30 Years of Historical Data

The deepest moat these companies have is not technical. It’s not code quality. It’s not UI.

It is decades of accumulated historical data.

This data is not just rows in a database. It is:
•20 years of regulatory changes
•20 years of billing exceptions
•20 years of edge cases
•20 years of audit trails
•20 years of institutional memory

It is literally the record of reality for that organization.

⸻

A Concrete Example: Municipal Finance Software

Imagine a German city that has used the same municipal finance system since 2003.

That system contains:
•property taxes
•business taxes
•water and waste billing
•fines and penalties
•exemptions and payment plans
•every correction, reversal, and adjustment

This is the city’s system of record for money.

Now an AI startup shows up in 2026 and says:

“We have an AI-native platform. Better UX. Cheaper licenses.”

They can absolutely build the software.

But replacing the old system means:
1.Migrating 22 years of messy data
https://t.co/IpO4e4KYdq and interpreting undocumented edge cases
3.Mapping everything to a new data model
4.Reconciling historical balances
5.Re-auditing the entire system
6.Getting regulators to approve it

This is a multi-year, multi-million euro project.

At the end, the city realizes:

“We will spend €5–10 million, take 3–5 years,
and risk legal and career-ending mistakes,
to save maybe €80k/year in licenses.”

So they don’t switch.

Not because the new system is bad —
but because the old data is too valuable and too dangerous to move.

That is the moat.

$CSU.TO $ACP.WA $SGN.WA

[Patrick Fagen @patfagen]

@gaurdipak Would love to get an update from Karl as this is 2.5 years old now.

[Andreas Steno Larsen @AndreasSteno]

I enjoyed this take

[Patrick Fagen @patfagen]

@sweatystartup It’s getting vibe coded now.

[Patrick Fagen @patfagen]

@sandeep_PT Surprised you didn’t touch on the shift these AI bros have made from ex: (Failed) Crypto & (Fraud) ICO, to ex: (Failed) Web3, to ex: (Failed) NFT, to ex: (Fraud) SPAC to AI advisors. It’s insane.

[Patrick Fagen @patfagen]

@brentmuio

[Rachel @tolstoybb]

First vagueposting induced bear market.

[Ed Zitron @edzitron]

this whole story is"researchers and investors who are directly invested in this are saying vague things about how powerful it is, and also a confirmed scam artist's AI-generated slop post went viral," with a smidge of "the companies are also saying the models are powerufl"

[Patrick Fagen @patfagen]

@PythiaR @4drant Prob means rates went down.