Testing AI’s GeoGuesser Genius [Scott Alexander/Astral Codex Ten] – A detailed walkthrough of how good general purpose AI [as of May 2025] is at guessing locations based off a single photo (stripped of geo metadata of course). It’s fun to try and understand how the models reason the answers here.
When Software Buys Software [Jeff Morris Jr/New Internet] – Early examples of what happens when AI models start to recommend certain software infrastructure products in response to general questions, and even more so, what happens when they make their own ‘purchase’ decisions agentically.
“Software won’t be sold over drinks. It’ll be selected, evaluated, and integrated by agents.”
One of the reasons everyone is trying to figure what SEO looks like in AI-world (coined AEO for now, done by companies like ScrunchAI, one of our recent investments).
In Switzerland’s member-based cooperative housing, new residents buy shares to gain admission to the building and get one vote in the corporation regardless of how many shares they own. The co-op uses the money to maintain the building, keep rents below market rate and, often, provide communal amenities like child care.
When a resident moves out, their shares are returned at face value. There is no capital gain.
Addition vs Subtraction [Molly Graham/Lessons] – Positing that organizations are inclined to always add more to their plates than subtract. And why that’s a problem. Five tools from Molly to try and get better at subtracting.
The reliability crisis [BuildingSlack/Johnny Rogers] – “If we can’t ship safely, we aren’t shipping at all.” That’s what Slack’s OG CTO told the company in 2018 when the product started to be plagued by reliability challenges. Company histories tend to be hagiographies or written by the most political of the team, looking to enshrine their own role in the success. BuildingSlack is a fun newsletter that sporadically fires off stories about the startup, written by former employees. This particular edition feels familiar to anyone who has ever gone through hypergrowth company phases.
A Fungible Worldview, Why Cluely Is The Dark Spirit of Venture Capital, Whether You Like It Or Not [Kyle Harrison/Investing101] – Typically I’m skeptical when an investor criticizes a startup – focus on your own house so to speak, unless there’s egregious behavior to be called out. But here Kyle uses Cluely – and his distaste for their stunt marketing – not so much to harp on them. Instead he shifts most of his focus back to the world he lives in – venture capital – and why Cluely is a chicken (or egg) example of Venture Capital 2025. And I think ultimately his post shares more about *him* than it does attempt to take down anyone else. So that I applaud. Plus I enjoy Kyle.
What Happens When AI Schemes Against US [Garrison Lovely/Bloomberg] – I live in the future. How else would I, in 2016, be able to write this post about commerce soon being agents trying to convince us to buy stuff, and eventually our own agents arguing back? Self-congratulations aside, I love the analysis here about current LLMs optimizing for a task (or ‘winning’) in a way that ends up potentially causing harm. I mean, this is basically the ‘what if AI decides the best way to ensure peace is to kill all humans’ question but not increasingly with real data. And hopefully lower stakes.
Startup org design: Design power centers intentionally – So good. Instructions on how to think about your org chart as power centers and not just resource management. As Molly says, “One of the most powerful things you can do to accelerate your company’s growth and reduce wasted energy is to design your org carefully including where you want the power to sit.”
Part 3: Compensation for Startups: Implementing and “Defending” Your Compensation System – Where Molly makes the case that ‘simplicity’ is a key factor of comp system quality and sustainability. “When you’re designing and implementing this system, you have to realize that a lot of the day-to-day will be implemented by recruiters and managers through hiring and through conversations with employees. That means that simple is your friend. As you’re designing, you want to think about the most junior recruiter on your team and ask, “With a little training, can that recruiter maintain this system?” The more complicated you make it, the more likely that the answer is no.”
Every marketing channel sucks right now [Andrew Chen/a16z] – A guy who made his career in growth marketing surmises that, yeah, it’s not just you, it’s everyone
Unfortunately this is the state of growth marketing. A lot of channels are not working, or are slow, expensive, or one-time only. This is the natural end state for things, and maybe we’re in a bit of a lull due to the technology super cycle as we’re 15+ years into the mobile wave, we’ve had various kinds of paid ads for 20+ years, and so on.
Andrew does provide advice though on how to navigate this stage but be forewarned, in his eyes a great product is necessary but not sufficient.
What was Quartz [Zach Seward/fmr Quartz CEO/owner] and Fell in a hole, got out [Tony Stubblebine/CEO Medium] – Pairing these because they’re both pretty honest looks at content businesses, the former a media company of some sort, the latter a publishing company of some sort. I append ‘of some sort’ not to degrade or minimize, but because the struggles and journeys they describe involved straying from, examining, and trying to return to, the mission and structure which made sense for the company. Quartz is largely a story told, while Medium is a story ongoing.
The most San Francisco sport ever? Underground robot boxing enters the arena [Zara Stone/The San Francisco Standard] – Please don’t turn this into a startup, a sponsored recruiting event, or #content. Against the backdrop of AI boom, our national government, and polarization generally, it’s fun to have something that’s just Keep San Francisco Weird.
It costs a US Hampton franchisee less than $5 per occupied room to furnish this cornucopia, but to a family of four, the perceived value is closer to $50, or roughly one-third of the average cost of a nightly stay. That math has helped power Hampton Inn’s unlikely rise to become the world’s largest lodging brand, with almost 350,000 rooms spread across 43 countries.
Obviously that’s not the entirety of the strategy – and you should read the whole piece – but I love the 10x ROI on waffles anecdote.
Encore Anxiety [Anu Atluru/Working Theorys] – Wonderful read on the pressure to succeed a second time. Oh please, some might concern troll, but I’ve witnessed this in tech careers. Most frequently with (a) repeat founders who fear their next company will fail and prove their initial win was a lucky fluke, and (b) younger folks who ending up joining an early stage rocketship out of school and once it succeeds beyond their expectations, worrying they’ll never find another. As Anu writes,
Impostor syndrome gets all the attention, but encore anxiety is its cruel foil: not the fear that you’re a fraud, but the fear that you’re genuine and still might not be able to prove it. The difference in attribution matters to the dominant psychology at play: the impostor fears their past success was luck; the encore-anxious person believes it was skill and yet fears they can’t summon it again at will.
Either way, there’s a focus on what others are thinking.
Ultraviolet Catastrophe: AI is about to make the public internet useless [Philip Rosedale/Philip’s Newsletter] – You might assume this post is about business model incentives pushing quality content behind paywalls or disincentivizing its creation all-together, but it’s actually about trust and malignant AI content (fraud, slop, etc). Philip believes the volume will be so significant that we’ll need to rethink the architecture of the web itself. His summary paragraph:
AI is about the flood the internet with messages and render it useless for many tasks. To scale to billions in a world filled with AI agents capable of typing 1000 times faster that us means a complete overhaul of our aging internet architecture. We will have to turn off anonymous public services, replacing addresses with channels – or in the language of graphs – replacing nodes with edges. There will be turbulence during this process, so fasten your seat belt and find some good books to read during the down-time.
Silicon Valley Etiquette [Angelo/Parallel Lines] – This was a really interesting sociological view of SV etiquette and the history of these norms. I hope we’re able to continue keeping the best aspects of our community, while reexamining how we also unintentionally (and intentionally) create barriers.
The Editorial Battles That Made The New Yorker [Jill Lepore/The New Yorker] – Skip this if you’re not a writing/creative process egghead like me. But for those who are, this is a great tour through the New Yorker’s editorial philosophy history and how it conflicted with (or helped) the writers they contracted for articles.
Inside Home Depot’s $20 Billion Secret Garden [Ben Cohen/Wall Street Journal] – Home Depot doesn’t just stock flowers and plants, it very much works hand-in-hand with growers, shaping what gets grown, what genetic attributes to optimize, and ultimately, what gets brought to market.
To find those plants, Home Depot runs 25 trial gardens in nine climate zones across the U.S. and studies them in the field under a variety of conditions. After all, a plant that thrives in New Mexico might not survive in New Jersey. For security purposes, some of those experimental gardens are hidden in cornfields or through backyard donkey corrals, protected on secret farms before the plants are selected and patented.
A Note on the State of Applicant Fraud [Matt Hoffman/M13] – Recruiting is changing dramatically because of AI – the LLMs now polish resumes, generate perfect responses, and in some cases, even create the candidates themselves. Some early data and how actually AI might also be the solution from Matt Hoffman, VC firm M13’s talent partner.
Whoa, it’s been a minute. Somehow went almost two months without a post. Bad blogger!
Yesterday Satya and I had the chance to celebrate Chime going public on NASDAQ. There are many ways to ‘grade’ an IPO from a punditry perspective (here’s my own framework which I think still holds), so I’m not going chime in with my own speculation (see what i did there), but on the other side of it, did want to recognize why this one felt special. It was our fourth IPO as a firm (defined in the loosest terms as Homebrew holding pre-IPO stock in a company), but the first which checked off all three of the boxes of consequence.
Financial Outcome for the Team: Founders and VCs know how to make money over the course of a journey like this, but what makes me giddy is seeing so many people on the Chime team win too. Being part of a special startup pays off in terms of career trajectory, learnings, and network, but also seeing your equity convert can change your life. I’m sure $CHYM minted many paper millionaires, but regardless of amounts, getting some dollars into the bank account allows these folks to continue taking career risk, which is positive for the ecosystem.
We Were There Since the Beginning: DPI without stories is boring. Stories without DPI is fatal. Chime provides us both DPI -and- Stories. The best combo! Special credit to my partner Satya who sat on their Board for many years, interviewed most of the early hires on behalf of the founders, and generally did more work than me
Founders are Still Running the Company: To have Chris Britt become the CEO of a NASDAQ company is a credit to him, and a role model for our industry. Chris came from a working class community in the New York City suburbs and brings that mentality, that respect, and that ethic to every day on the job. I’m romantic (and old) enough to believe going public means something.
Onward. Not the end of our relationship with Chris, Ryan, and Chime. Just the beginning of a next phase.
Dec 2024 drinks with Chris and Ryan, toasting what was planned for 2025…..
Soon I’ll have spent more time on cap tables than org charts. That’s a 2025 milestone as Homebrew turns 12.5 years old, surpassing my combined working tenure across Second Life, Google and YouTube. I entered venture capital with some beliefs – many of which still hold true (such as ‘your LPs are your business partners, not your customers’). But I’ve also seen a few change quite dramatically based upon the progressing ‘game on the field’ and my own VC experiences. One example is whether it’s assumed that seed VCs maximize outcomes by religiously holding their shares until the company itself exits. I mean, we’re investors, not traders, right? You’re told ‘illiquidity is a feature, not a bug’ and ‘let your winners ride.’ But when the physics of the model shift, you often need to with it. [While I’m going to focus on investor secondary here, I support common share sales as well – for example, back in 2014 writing “Getting Some Founders Early Liquidity Can Benefit VCs” during a period where many founders were being shamed for even asking about taking some money off the table.]
Ok, so what has changed by opinions about seed stage and secondary and why will the best early stage investors know when to sell, not just when to buy? Here’s the logic underpinning why ‘buy and hold’ is being replaced by ‘buy and maybe sell.’
Was
Now
Impact on Early Stage
Timelines to Startup Exit
On average 7-10 years to IPO, M&A
10-12 years+ as founders want to keep companies private; narrative that ‘bar is higher’ to go public; more grow/crossover capital to support private companies; periods of slower M&A due to private company valuations and/or regulations
Delayed liquidity hurts LPs who manage to an IRR and even for Cash-on-Cash returns slows distributions which can be reinvested in VC and other classes
For the earliest funds (pre-seed, seed) this means instead of 10 year fund cycles for LPs, you’re seeing closer to 15, which fundamentally changes LP calculations about the asset class
CoInvestor Alignment
Mostly structural alignment across the venture sector. Everyone largely underwriting to the same outcome goals.
Growth investors were the ones who added structure to deals and best companies typically just raised a single growth round ahead of IPO.
The dominance (in scale) of the multibillion dollar AUM holders, who are often underwriting to lower outcomes and needing to put more capital to work. That is, they rather have a 5x with $300m in the company than a 10x with only $30m invested.
The alignment gap between investors *starts* at the Series A, meaning earlier preferred investors cannot assume their interests are always aligned with the rest of the cap table. Angels and seed investors are better off thinking of themselves as common with a 1x preference once tens and hundreds of millions of dollars have been raised by a company.
How Investment Rounds Are Priced
Price discovery and valuation by within a relatively small community, with an independent new investor setting market price
A global auction filled with investors who have all sorts of objectives, experience, and return goals
I’m not bemoaning higher prices – the market bears what it bears and founders will make the decisions that they believe are best for their company. But this dynamic, for certain classes of companies, also means that startup pricing is often enthusiastic, optimistic and gives the company ‘credit’ for execution against forward looking plans quarters or years into the future. This decreases the penalty of ‘selling early’ to seed investors, and adds more performance risk to the investment, especially when seed investors lack the capital to protect/recap the company.
GP Incentives
You get really rich off of carry
With megafunds, you get really rich off fees regardless, which can impact all sorts of incentives to keep private marks high (TVPI!) while you raise new funds.
No Tiny s needed, but for more modestly sized pre-seed and seed funds, the returns are where you hope to strike it rich. So DPI matters sooner.
Infrastructure Around Secondary
Opaque, shady
Several large market makers, investor and company counsel have seen this before
There are now standard and trusted processes that reduce risk for all parties around these sorts of transactions. Still need to be careful working with unscrupulous parties.
Impact Upon Startup
Any VC selling is a warning sign that something must be wrong with the startup because they have inside info
Sure, there are cases where this might be true, but increasingly, and especially when the shares are bought by other existing investors/sophisticated players, it’s less of a concern
Balancing and consolidating the cap table on behalf of the founder to make sure the later investors have enough skin in the game. Sometimes we’ve seen founders proactively asking if we want to sell because they have more investor demand than they want to service.
VC Skillset
VCs are investors, not traders. We hold until the founders and company exit.
VCs increasingly *are* traders. Every venture firm who has held crypto tokens/coins have made buy/sell decisions and some even have a trading desk equivalent.
YOLO, not HODL
Now, optimally the secondary sales will always occur with the support/blessing of the founders; to favored investors already on the cap table (or whom the founders want on the cap table); without setting a price (higher or lower than last mark) which would be inconsistent with the company’s own fundraising strategy; and a partially exited investor should still provide support to the company ongoing. But even here I recognize than in some extreme situations you, as an investor, are forced to make calls about divergence in needs between your own, co-investors, and founders. The question is can you do it professionally and situationally enough to not harm the company and not develop a reputation for being a pain in the rear. As an industry peer said to me, “I think friendly secondaries are easy, everything else feels new.”
A second point of clarity is often the secondary is being performed for reasons other than just distributions to LPs, but also helps the venture firm recycle capital to support other startups in the firm’s portfolio. That is, early partial liquidity isn’t solely about investor wealth capture but is *good* for other founders in the ecosystem. Cash flow for small firms in pro rata, bridge rounds, and so on is a real challenge, and it impacts young startups disproportionately.
For funds like his, selling stock of private startups to other investors will be “75% to 80% of the dollars that [limited partners] get back in the next five years,” Hudson told me from his office in San Francisco’s brick-lined Jackson Square.
and
Hudson said majority of the capital he’s returned to LPs over the past few quarters was through secondaries, but declined to give specific names of the companies he sold.
And my former Google colleague, turned VC Tomasz Tunguz recently wrote a data driven analysis which concluded “It’s [secondary sales] not just a temporary anomaly, but a structural evolution in how venture capital will function.”
And trust me, there are many more who prefer to keep this type of activity private but are active harvesters. Secondary is quickly becoming primary for early stage VCs.
s needed, but for more modestly sized pre-seed and seed funds, the returns are where you hope to strike it rich. So DPI matters sooner. 