Oz Nur

@JaredSleeper the skepticism on AI ROI / HS capex concurrently with death of software is another odd one. one thing on the neoclouds: they’re the most reflexive stocks to express supply crunch but their burn / balance sheets make them scary.

Geniunely curious- can someone explain the bifurcation between Silicon Valley mega-bullishness on inference spend/token volumes and the softness in inference-levered public equities? Coreweave, Nebius, NVIDIA, etc. It is a pretty remarkable incongruence.

This efficiency point is one I’ve been stuck on too. You have to go company by company (to your point, Anthropic would stand out from any perspective). However, a meaningful percentage of AI-native growth trajectories are underpinned by compute pass-through (inference businesses) or token pass-through (applications) that ultimately point more to insatiable demand for the underlying infrastructure than to durable standalone economics. There is a bull case where the cost curve improves and AI natives create more value on top, but for the time being we should be cautious about touting “xx-employee company reaches xxx ARR” when the gross margin profile is masking how much of that revenue is effectively a conduit for compute or token consumption. There are rhymes with infrastructure software, where early negative gross margins were a natural feature of scaling (when, not if) — but that analogy can be applied imprecisely to many of the companies we see today. ARR per FTE still means something, but it doesn’t mean exactly what it used to — in a world where revenue can scale on the back of pass-through compute and token demand, the metric captures velocity without fully distinguishing between genuine operating leverage and infrastructure-subsidized growth. I’m sure it will evolve to carry more signal, but it’s hard to overlook that many of these trajectories are riding voracious underlying demand for compute and tokens, often at negative margins.

Yes, the public guys are an embarrassment at this point. That said venture capital was huge contributor to this bloat. Their hands are all over software and the culture of compensation/creation. Endless new companies founded w hacks to faster growth and faster ipo all grown within a managed environment. Now, as you pointed out, venture are more concentrated than they ever have been in their history in a few names. Basically riding shotgun w big tech in OpenAI/anthropic etc. Which begs the question on the ai doom loop… If companies are encouraged to radically fire in anticipation of ai handling more of the work, ai companies see their value go up the most. Convenient right ? But most white collar work agents will replace human wise is not complex and will likely consume minimal compute in a few years. So, then what? This is like the prediction machine in Ben afflecks paycheck film…

Spent the last few weeks pulling together our thoughts on the state of the software and ai market ahead of our annual meeting yesterday. Obviously a dynamic time with a lot going on so tried to unpack what is happening and our view on the different levels of risk and opportunities. Thank you to my colleagues @AdilBhatia and @lydianday for the hard work on this with me.

@andrew__reed catches me off guard when you toss in an insightful post in a sea of dad jokes and niche humor

the way to build shareholder value is to build a workplace that brings out the best in people. companies that focus excessively on might-be shareholders at the expense of their teams will bleed talent at the moment in time when talent matters the most. it’s unfashionable to talk about culture, and near uncouth to bring up the “work environment”. while surely the cold-brew-on-tap office perks of the 2010s were useless, and stock comp may be frittered away in uneconomic ways, the returns to building camaraderie and a winning locker room have never been higher than they are today. the best CEOs are building their companies both for their customers and for their teams. the shareholders will take care of themselves.

bam adebayo 😭

@levie @tszzl this is my style as a VC 😂

@tszzl A good 50% of my management style is “hey did you see this tweet, what are we doing about it”. My algorithm is *locked* in.

not sure what happened but in between last year and now twitter is back and not unusable slop

@FinnMurphy12 @tylercowen a breath of fresh air in a world of podcast slop

We aren't close to being prepared for the technological changes coming our way. In our first Forecast 2050 episode, @tylercowen lays out his greatest concerns for the next 25 years: - Collapsing birth rates - Aging societies + mass immigration - AI taking over everything by trick - Technology creating new problems - Why humans aren’t ready for the new era of rapid change Extended Show Notes & Transcript in the comments👇

@illscience @ekuyda phenomenal point

thanks ivan! high agency people tend to overestimate other people's interest in productivity - it's why agents won't kill apps @ekuyda is the one who pointed this out to me, and now I see it everywhere ....

chart crime

@jasonlk you’ve articulated my feeling well. great post

when the customer cube is already setup in quarters >>

@artlevy congrats!!

Largest bank-fintech deal of all time.

@JaredSleeper phenomenal work sir!

At Avenir, we’ve followed the emerging bear cases on SaaS closely. This 46-page deck contains our reflections and research on the path ahead. We see opportunity and risk as SaaS companies vie with AI natives to be "systems of context." Link in replies, excited to discuss!

If you feel stressed out, consider that you know too much information second-hand. About places you've never been, regarding people you've never broken bread with. Ground yourself with real work, done with your hands, and get some truths that you discover on your own.

@JaredSleeper such a good take. never articulated this well but 100% my pov

This is why it is very important for investors to be well-adjusted people who derive meaning from family, friends, hobbies, religion (if applicable), etc. and not just status games. Often, the market will reward (for years!) what is fundamentally low integrity "investing".

@martin_casado curious if you think this type of research furthers how far we can go with transformers in more novel sense or if it creates a relative improvements in the way the way we already use them?

Really incredible. If this holds up, it is a full and complete story of how LLMs work. Tl;dr transformers converges on Bayesian inference where you can predict the posterior function exactly. Excellent work by Vishal.

the older i get the more i've come to believe that, outside of luck, duration and focus are the main drivers of success. do things that compound and try again every day.

@bhalligan peter gassner of veeva. particular in a moment of narratives around systems of record, death of software, vertical versus foundation. probably one of most interesting POVs from last shift.

Who are some non-obvious CEO guests I should have on my pod?

@endowment_eddie most interesting spinouts from 2025 class have been Hanabi, Evantic, Chemistry. am excited to see what Victor Lazarte and Brian Singerman put together.

@oz_notes Give it a whirl

Top VC free agent/adds of 2025. In no particular order: 1) @yuris - General Catalyst 2) @RaghuRaghuram - a16z 3) @samteller - thrive Who am I missing?