Keith Smith

CEO & Founder, GoodStream - Real-time portfolio intelligence for private capital. 9 yrs building a private credit firm. Now focused on killing the PDF tax

[Keith Smith @GoodStreamSmith]

I spent 9 years watching private capital firms drown in PDFs, stale spreadsheets, and data they couldn't trust.

GoodStream fixes that. We ingest docs from inboxes and drives, extract and validate every field, and keep your portfolio continuously up to date.

7 active pilots, including Techstars. Launching widely in April.

Know every position, every day.

https://t.co/y9LFR2ZWfW

[Keith Smith @GoodStreamSmith]

@BoyanSlat Literally

[Keith Smith @GoodStreamSmith]

@theo Just wait till you find out about Siri.

[Keith Smith @GoodStreamSmith]

@satyanadella Great strategy. I wish all your software didn't suck.

[Keith Smith @GoodStreamSmith]

In 2000, Reg FD ended selective disclosure in public equities. Everyone called it a compliance cost.

By 2005 the verdict was in. Companies that leaned into disclosure got broader coverage, tighter spreads, cheaper capital. The ones that did the minimum got repriced.

Private capital is about to run the same tape.

Last month an operating partner told me it took 3 weeks and 4 analysts to answer one LP question: what is our true post-dilution ownership across 60 positions. The data existed. It was buried in thousands of pages of SPAs, side letters, and cap tables.

- 45 to 90 days from quarter close to LP letter
- 3 weeks to answer a question the fund should answer in 3 minutes

That is no longer a technology problem. AI reads the documents now.

So the excuse is gone and the advantage is open. The GP who can put a sourced, defensible number in front of an LP on demand wins the allocation from the GP who cannot. The founder picks the investor who can model the next round live in the meeting.

Nobody is forcing this. It is a rising tide with a scoreboard.

If your next LP asked for full look-through on the portfolio tomorrow, is that a Tuesday or a fire drill?

https://t.co/wVdp2Zy9YC

[Keith Smith @GoodStreamSmith]

In 2000, Reg FD ended selective disclosure in public equities. Everyone called it a compliance cost.

By 2005 the verdict was in. Companies that leaned into disclosure got broader coverage, tighter spreads, cheaper capital. The ones that did the minimum got repriced.

Private capital is about to run the same tape.

Last month an operating partner told me it took 3 weeks and 4 analysts to answer one LP question: what is our true post-dilution ownership across 60 positions. The data existed. It was buried in thousands of pages of SPAs, side letters, and cap tables.

- 45 to 90 days from quarter close to LP letter
- 3 weeks to answer a question the fund should answer in 3 minutes

That is no longer a technology problem. AI reads the documents now.

So the excuse is gone and the advantage is open. The GP who can put a sourced, defensible number in front of an LP on demand wins the allocation from the GP who cannot. The founder picks the investor who can model the next round live in the meeting.

Nobody is forcing this. It is a rising tide with a scoreboard.

If your next LP asked for full look-through on the portfolio tomorrow, is that a Tuesday or a fire drill?

https://t.co/wVdp2Zy9YC

[Keith Smith @GoodStreamSmith]

@DamienWayne Rich grown-ups buy tiny pieces of companies and then lose track because it all lives in messy papers. I built a robot that reads the papers and keeps a neat list of everything they own so nobody has to guess. https://t.co/lRNLNdrGDH

[Keith Smith @GoodStreamSmith]

@chamath @inverseOpinions Are you high?

[Keith Smith @GoodStreamSmith]

@chamath Because when has alcohol ever hurt anyone?

[Keith Smith @GoodStreamSmith]

@Geiger_Capital Crazy how many people are angry that poor people might get access to fresh food.

[Keith Smith @GoodStreamSmith]

@alexiskold I would agree if Mamdani was suggesting he take over grocery in NYC. Instead, he's acknowledging that in the outer boroughs, there truly are fresh food deserts - and is doing something about it.

This is food, not an existential threat to democracy.

[Keith Smith @GoodStreamSmith]

@harryjsisson Interior decorating and twitter trolling are his best and highest calling. I'd much rather him spend his time on this than on new wars and tariffs.

[Keith Smith @GoodStreamSmith]

@JayneZirkle It's good to see them practicing their pronunciation of his name. A few more rallies like this and they should be able to get it.

[Keith Smith @GoodStreamSmith]

@justinstorre @AnthropicAI And please do this about 2 years ago!

[Keith Smith @GoodStreamSmith]

I enjoyed my chat with Krista Tellefsen on the Business Credit and Financing Show. Check it out.

https://t.co/zT4TRqusNZ

[Keith Smith @GoodStreamSmith]

Ingestion, validation, and evidence used to be a competitive edge in private capital ops. In 2026, they are just the price of admission.

I was on a call last week with an operating partner at a mid-size fund. Her LPs no longer ask if she has a data pipeline, they assume she does. What they ask now is what she does with it, and how fast she can prove it when someone pushes back on a mark.

The market spent five years selling "we ingest your documents" as a feature. Then it spent two years selling "we validate the extraction" as a differentiator. Both are now baseline expectations, not selling points. If your platform cannot take in a cap table, a capital call, or a board deck and turn it into structured data without a human re-typing it, you are not behind the curve, you are not in the game.

Here is the part most people are missing. The real operating layer forming right now has three primitives, and they are sequential dependencies, not parallel features:

1. Ingestion: turning unstructured documents into structured data, fast and at scale.
2. Validation: catching what the extraction got wrong before a human sees it, using multiple models cross-checking each other rather than one model grading its own homework.
3. Evidence: showing, on demand, exactly where a number came from, down to the page and the paragraph.

Most vendors stop at step one and call it done. Fewer get to step two. Almost nobody has built step three as a first-class capability, and that is where the value now sits. Evidence is not a nice-to-have audit trail. It is what lets an LP trust a number without re-deriving it themselves, which is the entire point of paying someone else to do this work.

The second-order effect matters more than the primitives themselves. Once ingestion and validation are commoditized, the firms that win are not the ones with the best pipeline. They are the ones who can answer "prove it" in seconds instead of days. That single capability changes the shape of an LP meeting, changes how fast a capital call gets reconciled, and changes whether a GP spends the quarter defending a mark or building the next fund.

We built GoodStream around this sequencing, because we watched firms burn months on ingestion and validation with no good answer when someone asked where a number actually came from. The trifecta is not the differentiator anymore. What you build once you have it is where the next few years get decided.

https://t.co/y9LFR2ZWfW

[Keith Smith @GoodStreamSmith]

@libsoftiktok Imagine being so xenophobic that you confuse a city's greatest asset for its great liability.

[Keith Smith @GoodStreamSmith]

@skumWgmi We've somehow figured out how to tax the unrealized wealth of the middle class, but the rich keep tell us it simply can't be done to billionaires.

[Keith Smith @GoodStreamSmith]

@paulg Glee at the return of sanctioned institutional racism isn't the flex you think it is.

[Keith Smith @GoodStreamSmith]

Ingestion, validation, and evidence used to be a competitive edge in private capital ops. In 2026, they are just the price of admission.

I was on a call last week with an operating partner at a mid-size fund. Her LPs no longer ask if she has a data pipeline, they assume she does. What they ask now is what she does with it, and how fast she can prove it when someone pushes back on a mark.

The market spent five years selling "we ingest your documents" as a feature. Then it spent two years selling "we validate the extraction" as a differentiator. Both are now baseline expectations, not selling points. If your platform cannot take in a cap table, a capital call, or a board deck and turn it into structured data without a human re-typing it, you are not behind the curve, you are not in the game.

Here is the part most people are missing. The real operating layer forming right now has three primitives, and they are sequential dependencies, not parallel features:

1. Ingestion: turning unstructured documents into structured data, fast and at scale.
2. Validation: catching what the extraction got wrong before a human sees it, using multiple models cross-checking each other rather than one model grading its own homework.
3. Evidence: showing, on demand, exactly where a number came from, down to the page and the paragraph.

Most vendors stop at step one and call it done. Fewer get to step two. Almost nobody has built step three as a first-class capability, and that is where the value now sits. Evidence is not a nice-to-have audit trail. It is what lets an LP trust a number without re-deriving it themselves, which is the entire point of paying someone else to do this work.

The second-order effect matters more than the primitives themselves. Once ingestion and validation are commoditized, the firms that win are not the ones with the best pipeline. They are the ones who can answer "prove it" in seconds instead of days. That single capability changes the shape of an LP meeting, changes how fast a capital call gets reconciled, and changes whether a GP spends the quarter defending a mark or building the next fund.

We built GoodStream around this sequencing, because we watched firms burn months on ingestion and validation with no good answer when someone asked where a number actually came from. The trifecta is not the differentiator anymore. What you build once you have it is where the next few years get decided.

https://t.co/y9LFR2ZWfW

[Keith Smith @GoodStreamSmith]

@finkd Even if it sucks, it puts downward pressure on token pricing - which my $20k/month token habit truly appreciates