This is cross-posted from my semi-monthly newsletter. I'm the first (but not the last!) person to work full-time on the Neighborhood, an open collective of 200+ builders, founders, researchers, artists and more living within a single square mile in the heart of San Francisco. Our goal is to create an intergenerational campus full of radical agency, inspiring people, and unplanned encounters with friends. Check out past updates at jasonbenn.com.
It's been almost two months, so you know what that means: time for my monthly update.
To close the loop on the last update's cliffhanger: we did pick a location! We chose it unanimously and have only grown more confident in the two months since, which is helpful to me as I justify my slightly obsessive-compulsive choice to spend 300 hours researching it and consulting with a board of advisors. I want to post it here, but I’m going to refrain, if you’ll forgive me, because the area does have very limited housing inventory. But it's a bit of an open secret and I'm happy to tell you if you DM me.
We also made a landing page - sign up for our mailing list at neighborhoodsf.com. It includes bios of most of the people that have most helpful in coordinating this project so far. We'd like to rotate these bios on a regular basis, to better showcase our diversity.
Contents:
- Momentum
- Hyper-Local Real Estate Guy
- ⬇️ The part that wasn’t posted to Facebook ⬇️
- Challenge #1: rental dealflow
- Challenge #2: renter expectations
- The vibe is already coming online
- What’s next, and ways to help
Momentum
The Neighborhood has entered a slightly surreal phase where ~25% of the people we meet at parties have already heard about it. To be fair, sometimes that's because Kanjun got there first and already told everyone. But not every time!
We had our first low-key Neighborhood Meetup last Sunday. I asked everyone to divide into two groups - one for people that wanted to move to the Neighborhood in the next year, and another for people that were not. The groups were like 27 and 3. I’d thought they’d split ~50/50. Belated apology to everyone whose personal nightmare is standing up to introduce yourself to a group of 30 strangers.
I also made a major update these last two months. I wasn’t initially sure how I could be most helpful to aspiring residents. I figured I’d end up matchmaking people to housing opportunities. Turns out that when you live in a city full of hyper-competent people that they’re far better at finding places for themselves than I ever could be. Not to bury the lede, but our first group bought a 6BR in the Neighborhood a couple weeks ago, and another group signed a lease for a 4BR this weekend!
That people are so good at finding their own housing is frankly a big relief - it means I get to focus on all the other ideas we have for making the Neighborhood great.
Another pleasant surprise has been the Voltron of wholesomeness that is Patty Mou and Adi Melamed, who are joining forces to create a new third space in the Neighborhood called the Commons. If the term “third space” doesn’t mean anything to you, then check out Patty’s introduction to the idea - essentially, it’s a “community’s living room”. See also Patty’s Twitter thread about the first meetup.
Hyper-Local Real Estate Guy
You know, I had kinda hoped that after spending 9 years teaching myself programming, computer science, and then machine learning, I'd get work on... you know, all the mind-blowing shit now coming out of OpenAI, Anthropic, and DeepMind on a weekly basis. But I felt I had to quit my AI engineering job last June because I couldn’t shake the feeling that my marginal impact on AGI is pretty tiny compared to my marginal impact on long-term community. AGI is going to happen, with or without me, no matter how hard I work on it. A long-term community like the Neighborhood, on the other hand, probably wasn’t going to just appear. So I’ll pour all my energy into this for 2+ years instead, and we’ll see what happens.
Now, 9 months after that decision, it seems that the Neighborhood will be significantly better off over the long term if I become a hyper-local real estate expert. So fuck it, that's what I'm going to do.
Here's the situation: the Neighborhood is composed of renters and buyers, including folks living alone or with family, large coliving homes, DINKs in epic coliving mansions, third spaces that may or may not be independently profitable businesses, and "pocket neighborhoods", which are adjacent homes that merge common spaces (like backyards). Houses in the Neighborhood will be independently governed, because a decentralized structure creates both diversity and resilience. Fortunately, for most of these property types, people can procure their own housing. But some housing types - epic coliving houses, unprofitable third spaces, cheaper coliving houses, etc - we'll need to coordinate to procure. At least, unless there are any charitable billionaires reading this. Nah, just kidding. Go cure malaria instead, please.
One problem is that large homes are typically not available for rent. How many owners are going to try and fill a 12BR house with one shared kitchen? Basically none. But if we can buy these houses together, we not only get to de-gentrify our city by replacing some insular family with a culture-generating community house, but we also get to modify it over the years, making it ever weirder and more awesome. I want homes that become ART.
If we can make these opportunities legitimately competitive investments, we'll be able to raise orders of magnitude more capital from ourselves. This is the main reason for me to get into real estate. So expect to see more financial-themed posts in my future.
Rose always thought she'd marry a finance bro. Hilarious to me that she gets more hot and bothered by real estate math than AI math. Babe: I went from topology and variational inference to using FOUR BUTTONS on my phone calculator.
To that end, I’m partnering with my new sensei Ari Shpanya, the sponsor of the real estate syndicate that owns a couple community houses and dozens of others. He's already walked me through every acronym and formula in his pro formas, is giving feedback as I create my own from Redfin listings, and he’s down to partner on future investments, including taking on liability and using his own lender relationship to get as much as 85% LTV mortgages. Mostly, though, he's been trying to inoculate me as best as he can against all the bullshit and corruption of San Francisco's real estate business.
I also hired a realtor. This will be a long relationship, so spent 20 hours on it on it: I cold-called 15 agents with sales histories most densely concentrated in the Neighborhood's territory, interviewed them for general sales competence (Rose taught me how to evaluate this), alignment with purpose so I can trust their recommendations (thanks to Danny Hernandez for helping me develop the evaluation criteria), and real estate investment experience. After 12 phone conversations and two coffee interviews, I hired Marcus Lee from Compass. He's sold $500M in deals over 15 years, is a 5th generation San Franciscan, has personally invested in 25+ homes in San Francisco, and thinks the vision is awesome. He’s also open to co-investing with us, as long as the deal makes sense.
I’m also going to become a licensed real estate agent, lol. It only takes $800 and 135 hours of classes, and in exchange I get 0.625% of every home sale that I refer to Marcus. That pays for itself after just a couple sales. And 135 hours is nothing after 9 years of continuously learning computer science and machine learning. Little known fact about me, but flashcards is actually one of my love languages.
I’m not yet sure how difficult it is to go from newly licensed to successfully closing deals — agents encourage homeowners to go with buyer’s agents that they expect they’ll work with again in the future, which is mildly cartel-like behavior but whatever — but if this makes sense, then I could also potentially represent buyers myself and earn my income that way. Sales commissions are paid for by the seller, not the buyer, so my incentives would still be aligned with buyers. It’d be nice to have income again — I’ve got about 10 months of runway left and 100 months of ideas for things to do with the Neighborhood.
⬇️ The part that wasn’t posted to Facebook ⬇️
I'll quickly describe another couple problems that have emerged as the Neighborhood has progressed, and how we’re thinking about solving them.
Challenge #1: rental dealflow
We saw this one coming. It’s harder to find the unit of your dreams when you filter your search down to one square mile. This is a combination of two problems: inadequate dealflow and uncalibrated expectations.
My initial strategy to improve dealflow was about what you'd expect from a techno-optimist: solve the problem with programming! Rose loves to roast me for this, but whatever, this is me having fun.
I combined data about every house from San Francisco’s public data portal, the names and mailing addresses of every owner from a realtor friend’s private data portal, and then put on my detective monocle and found email addresses for a bunch of promising leads. Then I conducted an email drip campaign to the owners asking when their houses might next be available. Dear reader, this was an abject failure. I’ve since learned from realtors that drip campaigns are a common and ineffective practice, homeowners are tired of the spam, and response rates are <1%.
That said, my friend Janine, who’s building a smaller Neighborhood down in Austin, has had some success with handwritten notes to her neighbors that know her personally. Which brings us to our next experiment.
It turns out that we have a network of friends and allies that already live here. We conducted a series of surveys and asked around, and learned of NINETEEN intentional communities and 32 other houses full of friends that are already here. If every one of these people kept an eye out for neighbors that look like they're moving, how many off-market opportunities would we be able to identify? More than any realtor, that’s for sure.
In this world, the Neighborhood would act as a conduit and a routing layer through which off-market opportunities flow. This is how the Archive expanded, in fact: we started as the 9BR in 283 Dolores, snagged 281 when it became available two years later, 279 the year after that, and bid on 275 as soon as it hit the market the same year. Landlords fill their units faster, friends live closer to friends, hopeful tenants find rentals faster, and we'll start to organically agglomerate new Noasis-style pocket neighborhoods. It's also a nice incentive to befriend your neighbors! Most of them won't be moving for a very long time anyway - average turnover for owner-occupied homes is 5-10% per year, and 15-20% per year otherwise. We haven’t tried activating this network yet, but it’s already kinda yielding fruit by just discussing it with people - a friend told me last Friday that a 3BR will be available soon in his building.
Challenge #2: renter expectations
Realtors often complain to me that the set of homes that a client wants is frequently non-overlapping with the set of homes that are available and that they can afford. The solution? Hey, Rose — how about an app?
I was quarantined last month with Omicron anyway, so it’s not like I could do much else. So, after I recovered, I built an app on top of the dataset I described earlier that would allow us to rate every house in the Neighborhood from 1 to 10 (based on this rubric). Then, I scraped every Neighborhood house in Zillow (s/o to DK for building Browserflow), netting 2,100 homes with an average of 15 images and information about pricing history. Then I made an interface that laid out all the information you’d need in one screen and had keyboard shortcuts for rating. This being at least the third data labeling app I’ve ever built, I of course adhered to best practices: we rated every home once, every home scoring ≥ 5 twice, and every home with a rating disagreement of ≥ 2 a third time. I also built a leaderboard with rater calibration scores and threw in a live-updating progress bar.
Then I hosted a Neigbhorhood Rating Party. Sixteen friends (thank you!) assembled on Pi Day and we created over 4,000 ratings about 40 minutes. The leaderboard worked like a charm and my competitive friends were helpless to resist. Congrats to our top 3 raters: Rose (1000 ratings, -0.3 calibration), Patty (700 ratings, 0.3 calibration), and Jose (400 ratings, 0.1 calibration) for showing us that power laws really do show up everywhere (everyone else: 200 ratings).
Here are two applications that I think would be cool to build with this dataset:
- How long would I have to wait for e.g. a 6BR rated 8 or higher? Is it more likely to be 3 weeks or 3 months? How many of those even exist? Imagine filtering this interactive map view by your criteria, seeing pictures of comparable homes, and using that set your expectations:
- Hyper-local OpenDoor. Using price history information from ~1,000 homes over the last 10 years, we could train a model to predict when a home will be available for rent or for sale. I have a simple model that passes hand-crafted features into an MLP and outputs a distribution of the likelihood that a home will be available for sale in the future, asymptoting to 100% over 30+ years. So far, it has learned... that there is a bump in likelihood 12 months after a rental is taken off the market. That’s something!
Want to help? Let me know! Here’s the code and (soon) the data: https://github.com/JasonBenn/neighborhood
I consider these initiatives low-priority but fun. I also can’t wait until I’m a licensed real estate agent and they let me access the MLS directly...! Muahaha.
The vibe is already coming online
Delightfully, surprisingly, it increasingly feels like the Neighborhood is already coming alive.
I’ve now had four Sundays where I begin with Lisa Wehden’s Writing Collective from 10am-12pm, and then the entire day is taken up by spontaneous adventures with friends I encounter on the street, at friends’ homes, or at neighborhood third spaces. Here’s a group that ran into one Sunday morning, that then proceeded to bop around the Neighborhood until sunset:
We went to a tea house to catch up and joke about artificial wombs, went out to grab bagel sandwiches, met some friends for a picnic, walked over to a meetup at a local community house, then went back to the park for sunset. Some of us were able to walk home for a quick nap and then rejoin later.
Rose, when she experienced this for the first time, turned to me and said “I get it. I feel like a freshman in college again.” This is the exact type of moment that gives me the most energy and hope for this project, and the moment that I want to reproduce for as many aspiring neighbors as I can.
What’s next, and ways to help
Most of my time in this next phase is going to be spent getting a 6-9BR house with plenty of space for hosting, that is central, and with mostly best friends who are themselves great hosts. We’ll partner with a RE syndicate if it makes sense to buy, but we’re open to renting, too. We want this house to become another center of gravity for the Neighborhood.
If you want to help: thank you so much! Right now, I’m mostly looking to meet real estate entrepreneurs or people that you think are super smart and that have real estate expertise. We're gearing up to potentially buy houses, so now's the time to figure out if there's a business model here that’s more clever than just buying them the old-fashioned way or through a syndicate.
If you made it through this whole thing, mad props. I appreciate you.
Warmly,
— Jason and the rest of the Organizers