Tomasz TunguzModify

2024-12-20 08:00:00

Every year I make a list of predictions & score last year’s predictions.

Here are my predictions for 2025.

1. The IPO market rips. ServiceTitan’s success has revealed the retail & instititutional demand for high growth software. Stripe, Databricks & many others generate huge liquidity for VC funds. The pull from the public market & desire for AI drive M&A to 5 year highs, enabled by a laxer FTC M&A policy.
2. Google continues their surge in AI. They lept from no placement to top 1 or 2 on the OpenRouter rankings. They further advance their market share. Grok benefits from Elon’s position in government to become a viable contender with OpenAI & Anthropic.
3. Voice becomes a dominant interface for humans with AI as speech models are pushed on device & the accuracy/latency astounds. Voice produces text, image, & video. Why type? It’s the start of a generation of people who will never learn to type on a keyboard. (This one will be hard to grade!)
4. US VC investment remains roughly around $210-$230b, but VC fundraising increases by 20% as LPs invest some distributions into the asset class, in pursuit of AI growth.
5. Consolidation is the theme for the Modern Data Stack. Buyers look to standardize on single platforms as cost-pressures persist. More than $3b of M&A in the category is announced. Software & data engineering teams continue to fuse.
6. The first $100m ARR company with 30 or fewer employees is created. An AI native product coupled to an AI native team produces incredible market cap creation efficiency.
7. After years of declines, the US web3 engineering populations grows by 25% as the government embraces crypto & web3. This opens to the door to 30% more token listings & a successful consumer app built on a web3 stack.
8. Data center spending by hyperscalers eclipses $125b for the year as the AI race fuels demand for GPUs. Broadcom is the hottest semiconductor stock of the year.
9. Stablecoin supply increases 50% to $300b as more businesses adopt this payment mechanism for B2B payments. Stablecoin volume is greater than 3x Visa’s transaction volume.
10. Observability, SIEM, & Business Intelligence begin to use the same data lake. Usage-based pricing for many software companies creates a need for a single data lake. The data lake becomes the dominant data architecture across all workloads.

Here are my predictions from last year.

1. The IPO market remains closed through the first 6 months of the year. But a few mega issuances, especially Stripe & Databricks in the summer or fall, re-open it for others. The Fed cuts rates, which helps.

   Score: 0. Only 3 IPOs! Maybe next year!

2. M&A accelerates throughout the year. The anticipation of a rate change drives fear of target acquisition valuations. In the last two years, M&A has totaled about $49b & it surges to above $60b driven by AI acquisitions. PE becomes an important buyer of companies growing 10-25%, as it did in 2018, driven by lower debt costs.

   Score: 0.9. M&A was up $10b y/y but not so much by AI.

3. AI & data continue to dominate the funding landscape as founders & investors seek novel applications of the technology. A handful of companies achieve record-setting growth rates.

   Score: 1. AI funding is now half of all VC funding. Companies are attaining record growth rates.

4. The share of AI-enabled web searches approaches 50% of all consumer search as consumer behavior patterns evolve, especially on mobile.

   Score: 1. Most GenZ & Millennials default to AI search, & for many Google.com search, an AI Overview graces the top of the page.

5. The BTC ETF drives a resurgence in interest in web3 financing. The winter forced many companies to evolve from open-source projects to revenue-generating businesses. We see the first broadly successful tokens with dividends (likely outside the US). This innovation reinvigorates very early-stage IPOs. We also see more ARR-based web3 businesses achieving scale. Record inflows into tokens fuel all-time highs in Bitcoin, Solana, & higher performance L1s who offer better price/performance to market.

   Score: 1.0. BTC at $100k & the ETF is the fastest growing ETF in history.

6. US VC investment falls from $275b in 2022 to $200b in 2023 & sustains at about $200-220b in 2024 as LP interest in venture attenuates after the euphoria in 2020 & 2021. Valuations remain relatively steady except for AI businesses, which command a premium to market of about 10-25%.

   Score: 0.7. US VC investment totaled $210b in 2024. AI premiums are closer to 2-5x.

7. The discussion around AI regulation becomes a critical topic in the US during the election because machine-generated content exacerbates international meddling in US politics. But the overwhelming desire for the US to continue to lead the innovation wave it started creates safe harbors, the same provisions which enabled the web to flourish, are applied to AI.

   Score: 0.5. Yes, AI regulation was a hot topic. But the election impact didn’t materialize as expected.

8. Companies & startups in particular begin to report meaningful improvements in productivity from AI, reducing their headcount growth, but growing revenue just as much as projected. ARR per employee increases 10%, twice the decade long average.

   Score: 0.7. Microsoft, ServiceNow, & Salesforce all report productivity gains from AI measured in the hundreds of millions of dollars.

9. Data lakes become the dominant data architecture across business intelligence & observability workloads as more startups leverage Amazon S3 free replication. Cloudflare R2’s architecture for very large data sets drives a meaningful growth in its usage, predominantly for AI.

   Score: 0.5. The narrative in data has certainly become the data lake. The R2 architecture hasn’t materialized.

Total : 6.3/10

Down from 4.0/5.0 last year! But it was an election year so I’m cutting myself some slack.

2024-12-17 08:00:00

From the pit lanes of Formula One to the secretive world of commodities trading, from the championship poker tables to storytelling competitions, I enjoyed a wide range of different books this year. Here are my favorites :

1. Every Hand Revealed by Gus Hansen: A hand-by-hand narrative of Professional poker player Gus Hansen winning the 2007 Aussie Millions tournament. Hansen had a unique style at the time, aggressively defending his blinds.
2. The New Map by Daniel Yergin: Pulitzer Prize-winning author Daniel Yergin explores the changing landscape of energy, geopolitics, & climate change. He presages the Ukrainian conflict in 2019.
3. The World for Sale by Javier Blas & Jack Farchy: This book delves into the secretive world of commodity traders, the individuals who buy & sell raw materials like oil, metals, & grains. It exposes the immense power & influence these traders wield in shaping global markets & how they supported some failing countries, like Cuba & South Africa, during energy crises.
4. Seven Brief Lessons on Physics by Carlo Rovelli: I love to read books about theoretical physics because it’s at the limit of what I can grasp. Rovelli describes quantum concepts in an approachable way.
5. The Seventh Floor by Richard D. McCloskey: Spy fiction is my way of relaxing. I’ve read every John le Carre book & discovered the Golden Dagger award, given to the best spy fiction each year & have started looking for new ones . “The Seventh Floor” is one in a series about the CIA’s efforts to subvert Russian spies.
6. How to Build a Car by Adrian Newey: Adrian Newey revolutionized Formula One car design with a background in aerodynamics. He is the single most successful car designer in the history of Formula One, having won many championships. This is his autobiography.
7. The City & Its Uncertain Walls by Haruki Murakami: Norwegian Wood was my first Murakami novel, & since then, his magical realism has echoed Gabriel García Marquez. I’ve read every one of his books, & this is the most recent, which ties back to some themes he wrote about when he was just starting as an author.
8. Storyworthy: How to Use Stories to Sell, Stand Out, & Win by Matthew Dicks: Matthew Dicks is an award-winning storyteller. He won several awards, including one for a story about a magic fork captured by one of his students. It’s worth listening to if you have 15 minutes. He shared some tactical techniques for creating a story, including developing an elephant & using a backpack - some great mnemonics for improving your stories.
9. Chronicles of a Liquid Society by Umberto Eco: One of the most learned men & broadest vocabulary, Eco is also famous for amassing a library of over 10,000 books. This collection of his essays laments progress. In other words, a curmudgeon’s foil to the techno-optimism I love.
10. The Dawn of Everything by David Graeber & David Wengrow: Graeber & Wengrow challenge conventional narratives about human history, arguing that early societies were far more diverse & complex than previously thought. It shares stories of Native American diplomats during the 17th & 18th centuries who traveled to Paris & may have influenced political thought.
11. Active Measures by Thomas Rid: The history of information warfare. Not a beach read, but fascinating.

Send me your recommendations for 2024!

2024-12-16 08:00:00

Remember the first time you touched a computer screen instead of typing commands?

We’ve lived through distinct epochs of human-computer interaction: the cryptic beauty of command lines, the intuitive dance of graphical interfaces, & the ubiquity of browser-based computing in the SaaS era.

It’s different now. When I manage spreadsheets, I don’t want to manipulate formulas anymore. Instead, I want to instruct the computer as I would explain to a colleague : “run the correlations on these variables to see if anything meaningful pops out, then plot it, & add it to the deck.”

When I parse research reports, I don’t want to read them line by line ; I want to ask for summaries & surprising conclusions.

When I program computers, I don’t want to work at the level of lines & functions & arguments, but the style of a webpage like this blog’s theme.

When I read an email, I want to command an AI to follow up on this in a week, or research the new prompt engineering technique in this newsletter & send it to my Kindle so I can read it that night, or compare the revenue multiples on health care records companies to horizontal CRM.

So, I’ve taken to keeping an AI open on a separate monitor. That’s step one.

But I’m starting to run models on my laptop so I can fire all kinds of questions at its feet & watch it dance. Can it draft an email, or critique a blog post well?

Running Llama3.3 on my computer is the nearly same as asking any of the major alternatives - but more private & a bit slower - which for the moment works for me. It buys me time to read the AI’s output and reflect on this new working relationship.

Talking to the computer using the transcription program I built (that I’ll soon share), reinforces the sense of collaboration.

I can see a world where I’m not flipping between applications. Instead, I’m telling the AI to send an email or research multiples & that’s my interface - that’s my OS. Not the command line, not the desktop, not the browser.

The AI choreographs all that work underneath.

2024-12-13 08:00:00

If you needed proof that every software company is an AI company, here’s the evidence. 46% of all US venture financings in dollar terms in 2024 were AI.

This revolution breeds in every corner of the startup ecosystem :

37% of Seeds, 35% of Series As, 67% of Series Bs, 24% of Series Cs, & 47% of Series Ds were AI companies. Within a year, half of software companies are AI companies.

By next year, if this trajectory holds, we’ll struggle to find software that doesn’t think, learn, and evolve.

What isn’t captured in this data is that the way companies are built is also changing. AI is in the product but AI is also writing software, crafting marketing company, & drafting design documents.

AI is more than a market trend - we’re watching the software industry undergo a cognitive reimagining of how we build software, they way we work & what is possible with software.

2024-12-10 08:00:00

“Feel Good Fun Mix” tops my recommendations on Spotify. Spotify has created over 6,000 such labels by hand.

Amazon and Netflix attributed 35% and 75% of their revenue to their recommendation systems. This is a profound & counterintuitive shift in how we think about marketing.

In a recent case study led by Aampe, an AI agent with this type of segmentation sent far fewer messages than traditional systems while achieving better results.

Even more striking, the AI agent learned to offer smaller discounts (15% vs. 35%) to users showing higher purchase intent, automatically optimizing for both conversion and profit margin. Streak marketing, urging users to read a few pages each day, for example, works wonders too.

Today’s marketers are limited in their ability to construct a vast number of user segments - for example, the midnight snacker who prefers to receive texts very late on a weekday isn’t an obvious cluster. A barber booking app found nearly 400 different user segments in their population.

Aampe is building the future of marketing - agentic marketing.

Aampe creates an agent for each user. The agent understands point of sale history (buying a coffee every morning at 745am), mobile app usage (does she pre-order before she arrives?), and engagement with previous messages.

By combining all this - ad engagement, surveys, & core business outcomes - Aampe produces effects on performance that are significant & sustained.

At the core of the product, reinforcement learning explores what a person likes - the message, the medium (email/text/notification/GIF), when they like to receive it & how these change with time - something not possible with the marketing canvases of today.

This emerging approach empowers teams to create personalized lifecycle journeys for each user across millions of users. They serve more than 50m monthly today. We’re announcing our partnership with Aampe by leading their Series A.

Led by a team of human behavior researchers at Harvard, US Army, & NATO, Aampe publishes their learning about agentic marketing here.

2024-12-09 08:00:00

How often do you use AI? I tracked my Sunday workday to find out. Between 4:30-9:00 PM (with a dinner break), I monitored every AI interaction while handling emails, analyzing data, & writing.

If the average American picks up their mobile phone 144 times per day & we call that addiction, I am using AI about a hundred times per hour. Is AI ten times more valuable than a phone?

Speech

Dictation activity is bursty with breaks for dinner. But looking at 7pm & later, I’m hitting the speech API at least once every 2 minutes but closer to every 90 seconds on average. Around 8:00 PM is when I started to email - massive activity spike.

On average, the typical transcription contains about 18 words, but more than a quarter of them contain more than 50 words. Those longer tracts are typically entire email responses dictated in one shot.

Dictation is 3 times faster than typing, so it saves me an enormous amount of time.

Two AIs process my voice : initially a dictation AI and a language AI that edits for brevity & clarity (both of these run on my laptop).

Coding

Publishing data-driven blog post analysis is a key part from formatting the data to analyzing it using R and then publishing charts. All of this is now predominantly handled with prompts to an AI.

I can generate several hundred lines of code in 5-10 minutes. With the newer models, I expect this to collapse to 1-2 minutes.

In fact, I find myself growing reliant on the AI to the extent that I no longer remember some of the R syntax, A sign of working at a higher level of abstraction : one of the promises of AI.

Fully focused on work, I’m employing AI roughly 50-100 times per hour.

Within the last 24 months it’s clear that AI has become an essential coworker, perhaps at least as important as a mobile phone, but very likely more critical.