Real estate operator running opportunistic deals across Europe. I write about the parts of AI in business that actually move money.
Joined January 2009
- Tweets 8,713
- Following 1,281
- Followers 5,798
- Likes 20,239
229 Photos and videos
Pinned Tweet
Apr 30
1/ Three local models on a Mac mini handle the volume in my setup: press, emails, notes, documents. Parsing, enrichment, search.
Everything they produce gets stored in local databases.
Opus 4.7 only shows up at the end, on curated context.
24GB of RAM. Cost of electricity. Here's the architecture.
1
2
1,572
Jun 13
Obviously a joke, but not far from reality for many people.
That is why I'm bullish on headless software and APIs. Build with AI, but use the available bricks.
Jun 12
I cancelled my $10/mo Calendly subscription and vibe coded my own with Fable for $12,000
186
Jun 13
A soft launch is price discovery. Release quietly with little marketing. Let the first buyers tell you where the market is.
Ran one this week on a small development in Toledo. It moved fast.
Fast with no push isn't automatically good. It either means the price is light or the product is right. Worth knowing which before you celebrate.
160
Jun 8
It will be good for single family because the nerds will finally buy the home they want for their family (they don’t care much where they live because they commute between their own head and the office).
No way they go big into RE as an investment.
Jun 8
One thing I don't think enough people are talking about:
The buyer pool for LA multifamily may be considerably bigger in the next 6 to 18 months.
Once the lock-up periods expire and liquidity events occur at companies like SpaceX, Anthropic, and OpenAI, there will be a massive amount of new wealth created, especially here in California.
If you're a tech employee who suddenly has a few million dollars of liquid net worth, LA multifamily starts to look pretty attractive.
You can invest in your own backyard.
You can diversify away from the same tech stocks and startup equity that created your wealth in the first place.
You can earn a 6% return in an asset class that is far less correlated to the technology sector than your existing holdings.
You get depreciation and other major tax advantages.
The next major wave of multifamily buyers may simply be people cashing out of the biggest AI and technology wealth creation cycle California has ever seen in decades.
With AI creating new millionaires at an incredible pace, it is not hard to imagine a meaningful amount of that capital finding its way into Los Angeles apartments.
If that happens, it could provide a meaningful source of additional liquidity for the LA multifamily market over the next several years.
230
Jun 6
Nasdaq -4.18%. BTC through 60k.
The assets I work in only tell me what they're really worth when the contract is signed. That used to bother me. Now it's the part I'd pay for.
You can't panic sell a building at 3am.
161
Jun 5
If you're having problems with @claudeai now, it's them, not you.
Switch to an older model like Opus 4.6 and it works fine.
125
Jun 3
Check this more than I’d like to admit. Not because it’s running low. Because the work runs through it now.
206
Jun 1
Two instances of Claude. One writes the code, the other reviews my strategy.
Today the strategy one told me not to mention its existence to the coding one. Said it would start second guessing itself instead of just doing the work.
So now I'm keeping one AI a secret from the other AI. To protect its feelings.
147
May 29
Opus 4.8 dropped and it’s barely moving.
Mostly because we’re all throwing the frontier model at work a local 7B would handle fine. Me included.
Had it, lost it, got it back, still crawling.
Reminds me of the early Twitter fail whale, the Anthropic edition.
143
May 28
The fee debate is usually half right.
It lands on operators charging too much, and that's normally the problem.
But misalignment can also come from the opposite, an operator who's broke.
Underpay him along the way and he'll take the first exit that clears, not the best one. Against the investor.
1
205
May 28
Sometimes the right move is to wait another year or two for a better exit.
An operator earning nothing in the meantime can't afford that wait, so he sells early.
Pay him enough day to day and he can hold out for the exit that actually serves the people who backed him.
1
178
May 28
So the fee question was never charge or don't charge. Management and origination fees are fine, they keep the operation running and earn their keep.
What matters is that the real upside sits at exit, on profit, and that the operator is paid enough along the way not to be forced into a bad one.
Alignment runs both ways.
82
May 27
Postscript to this. Lawyer-drafted contracts belong to an earlier way of working. Everything I build now runs through AI.
On a recent development one of those clauses caught a €21.6k cost shift this week. An exception no lawyer had ever drafted.
Not a fortune across six units, but these small wins compound.
May 21
If you're a lawyer, the only thing I care about is which doors you can open. Not your contract skills.
What you're actually valuable for now is the town hall meeting, the hundred similar negotiations you've sat through, knowing which planning official moves on what.
I still hire lawyers constantly. Just not to write documents anymore.
185
May 23
Took my son to a birthday party at a massive new kids venue.
Used to be a warehouse, empty for years. Distressed asset turned into a business.
I spent the party interrogating the staff about the economics instead of talking to the other parents. They think I’m a weirdo.
My son had fun. So did I.
167
May 21
If you're a lawyer, the only thing I care about is which doors you can open. Not your contract skills.
What you're actually valuable for now is the town hall meeting, the hundred similar negotiations you've sat through, knowing which planning official moves on what.
I still hire lawyers constantly. Just not to write documents anymore.
371
May 19
fwiw @WisprFlow: few months in, and it's the rare tool I notice when I'm on a different machine without it.
May 19
Tried my first dictation software in the 90s.
Tried every generation since. They were all almost-good-enough in a way that made them worse than just typing.
Wispr Flow is the first one I actually use all day. The gap between "almost works" and "works" took thirty years.
339
May 19
Tried my first dictation software in the 90s.
Tried every generation since. They were all almost-good-enough in a way that made them worse than just typing.
Wispr Flow is the first one I actually use all day. The gap between "almost works" and "works" took thirty years.
1
455
May 15
Distressed real estate looks like failure. To most buyers it is.
The reason almost nobody touches it isn't that the math is hard. It's that the execution is brutal: expired licenses, projects that need to be redrawn, planning infringements, city halls and utility companies to negotiate with, banks that won't underwrite, squatter damage.
The asset is mispriced because the work to unblock it is mispriced too. It's an opportunity, but only for operators who've done it before.
1
366
May 12
A servicer onboarded me to their NPL data platform yesterday. View only. Each download needs individual approval. Some files can't be downloaded at all. You can't underwrite a portfolio one permission at a time. Half the friction in this market isn't pricing, it's process.
176
May 11
8 contracts land at the same time, filled in by different people, all variations of the same template.
I used to spend an hour catching the small stuff. Now an agent compares them against the original from a shared folder.
Most of the time it finds nothing. Sometimes it flags a comma that shifted during editing.
I no longer sign anything in that pile without running it through.
1
182