the VC almanac

all things venture capital & tech

[the VC almanac @theVCalmanac]

> be 2009
> two guys decide to start a venture capital firm
> problem: every top VC has decades of reputation
> Sequoia backed Apple, Cisco, Google
> how do you compete with that starting from zero?
> one guy co-invented the web browser
> the other took a company public with $2M in revenue
> press called it "The IPO from hell"
> he sold it for $1.6 billion
> they raise $300M year one
> $650M year two
> $1.5B by year three
> start backing unknowns like Airbnb, Coinbase, Instagram, Slack
> every single one becomes generational
> Coinbase alone returned $7 billion to their investors
> by 2025 they captured 18% of ALL venture capital dollars in America
> January 2026: raised another $15 billion in one sweep
> now managing over $90 billion
> 1,076 portfolio companies
> more unicorns backed since 2020 than any other firm on earth
> oh and the CEO guy?
> started his career as a rapper
> runs a nonprofit giving pensions to overlooked hip-hop pioneers
> wrote the most famous product management memo in tech history
> out of frustration with his own team
> his life philosophy? "Life isn't fair"
> his firm? Andreessen Horowitz
> a16z

tl;dr two guys with zero VC reputation in 2009 built the most dominant venture firm in history by backing founders everyone else ignored. 16 years later they manage $90B, back 1,076 companies, and captured nearly 1 in 5 venture dollars in America.

[the VC almanac @theVCalmanac]

@hannaan_atomik How bro felt after orchestrating this:

[Hannaan Kirmani @hannaan_atomik]

Such a pleasure working on launching a company that's going to redefine the LLM space.

Congrats team!

[the VC almanac @theVCalmanac]

According to @dylan522p, hyperscalers are spending 50% of their CapEx on memory this year and we might not get new phones or laptops this year.

"Everybody's talking about the memory wall right now. Memory's getting super expensive. There's not enough memory. Smartphone volume will go down 30% because there's not enough memory."

"Hyperscalers are spending, this is shocking, Dylan said they're spending 50% of their CapEx this year. On memory."

"What is hyperscaler CapEx? It's like high hundreds of billions, maybe a trillion. And they're spending half of that on memory."

"That is a huge constraint. That's why we're not going to get new laptops and phones this year."

🎥 @dwarkesh_sp @reinerpope

[Dwarkesh Patel @dwarkesh_sp]

Did a very different format with @reinerpope – a blackboard lecture where he walks through how frontier LLMs are trained and served.

It's shocking how much you can deduce about what the labs are doing from a handful of equations, public API prices, and some chalk.

It’s a bit technical, but I encourage you to hang in there - it’s really worth it.

There are less than a handful of people who understand the full stack of AI, from chip design to model architecture, as well as Reiner. It was a real delight to learn from him.

Recommend watching this one on YouTube so you can see the chalkboard.

0:00:00 – How batch size affects token cost and speed
0:31:59 – How MoE models are laid out across GPU racks
0:47:02 – How pipeline parallelism spreads model layers across racks
1:03:27 – Why Ilya said, “As we now know, pipelining is not wise.”
1:18:49 – Because of RL, models may be 100x over-trained beyond Chinchilla-optimal
1:32:52 – Deducing long context memory costs from API pricing
2:03:52 – Convergent evolution between neural nets and cryptography

[the VC almanac @theVCalmanac]

AlexNet took 6 days to train in 2012. On today's NVIDIA chips, the number will shock you.

"AlexNet was a 60 million parameter deep neural network, and it was trained for six days on two GTX 580s, which was the top consumer card at the time."

"I was looking up some numbers last night just to put these in perspective. The newest, the latest and greatest from NVIDIA is the GB200."

"How much raw compute factor we have between the GTX 580 and the GB200? It's in the thousands."

"That training run of six days on two GTX 580s. If you scale, it comes out to just under five minutes on a single GB200."

The breakthrough that started deep learning now fits in a coffee break.

🎥 @jcjohnss @theworldlabs @a16z

[the VC almanac @theVCalmanac]

Why Every LLM still tells the same joke from 4 years ago?

"Apparently, the joke all the LLMs love the most is: why do scientists not trust atoms? Because they make everything up."

"This is the joke you would get three or four years ago. And this is the joke you still get today."

"Even though the models have improved tremendously, it's part of the jaggedness. Shouldn't you expect models, as they get better, to also have better jokes, or more diversity of them? It's just not being optimized. It's stuck."

Models got 1000x smarter, and they forgot to train their models on how to say jokes.

🎥 @karpathy @NoPriorsPod @saranormous

[sarah guo @saranormous]

Caught up with @karpathy for a new @NoPriorsPod: on the phase shift in engineering, AI psychosis, claws, AutoResearch, the opportunity for a SETI-at-Home like movement in AI, the model landscape, and second order effects

02:55 - What Capability Limits Remain?
06:15 - What Mastery of Coding Agents Looks Like
11:16 - Second Order Effects of Coding Agents
15:51 - Why AutoResearch
22:45 - Relevant Skills in the AI Era
28:25 - Model Speciation
32:30 - Collaboration Surfaces for Humans and AI
37:28 - Analysis of Jobs Market Data
48:25 - Open vs. Closed Source Models
53:51 - Autonomous Robotics and Atoms
1:00:59 - MicroGPT and Agentic Education
1:05:40 - End Thoughts

[Maker @makermember]

@acenario and @abhijay_cloaked just raised $375M for @keepitcloaked to fix the internet's biggest privacy problem.

In this episode, they take us back to the beginning and share the untold stories about everything it took to get here.

Episode 04 drops tomorrow.

[the VC almanac @theVCalmanac]

@saranormous cc: @steipete

[the VC almanac @theVCalmanac]

More about "Dobby the Elf" with @saranormous.

"I used the agents to find all of the smart home subsystems on the local area network. I just told it I think I have Sonos at home, can you try to find it? It logged in and said oh yeah, you have these Sonos systems installed."

"It did some web searches, found the API endpoints, and asked me if I wanted to try it. I'm like, whoa, you just did that. Can you try to play something in the study?"

"It did the same for lights. It hacked in, figured out the whole thing, created APIs, created a dashboard so I could see the command center of all my lights in the home. Then it was switching lights on and off."

"It controls all of my lights, my HVAC, my shades, the pool and the spa, and also my security system."

Video: @NoPriorsPod

[Andrej Karpathy @karpathy]

Thank you Jensen and NVIDIA! She’s a real beauty! I was told I’d be getting a secret gift, with a hint that it requires 20 amps. (So I knew it had to be good). She’ll make for a beautiful, spacious home for my Dobby the House Elf claw, among lots of other tinkering, thank you!!

[the VC almanac @theVCalmanac]

Alex Karp just warned Silicon Valley: you're about to get nationalized.

"If Silicon Valley believes we are going to take away everyone's white collar job — highly educated people who went to elite schools, who vote for one party — and you're gonna screw the military, if you don't think that's gonna lead to nationalization of our technology, you're retarded."

"And you might be particularly retarded because you have a 160 IQ. But this is where that path is going."

"You cannot have technologies that simultaneously take away everyone's job and then be perceived as screwing the military. The danger for our industry is that you get a famous horseshoe effect where there's only one thing people agree on — and that's that this is not paying the bills and our industry should be nationalized."

The $18B CEO of the company that built the Pentagon's intelligence systems is telling Silicon Valley point blank: work with the government or lose your company.

[the VC almanac @theVCalmanac]

Palantir CEO Alex Karp on the best outcome for AI in America:

“We show our adversaries: you can’t fuck with us.”

Via @tbpn

[Jawwwn @jawwwn_]

Alex Karp rips into the Palantir conspiracy theorists:

“You're attacking the person who's protecting you— idiot.”

“You may hate this, but there's one person protecting your rights to be a conspiracy theorist that actually has a seat at the table, and that person is me.”

“You may not want to hear that truth, but it's fucking true.”

“Maybe do a little more reading before you pontificate on your absurd and obviously ill-formed and many times stupid opinions.”

“It's like fucking so stupid.”

Via @tbpn

[the VC almanac @theVCalmanac]

Alex Karp: "We should publicly humiliate people who don't support our war fighters."

"We have American war fighters on the battlefield, willing to sacrifice their lives. Some of whom have sacrificed their lives. They have families and kids. The kids and families don't know if their loved ones are coming home."

"People who are not aware or somehow so spoiled that they don't realize what these people do for us, we should publicly humiliate them. Everyone has a role in that."

"Especially those of us who are riding a crest of intergenerational cultural, intellectual, courage advantage, which it is to be an American nowadays, should not forget the war fighters who are disproportionately from the middle of our country and disproportionately have gotten screwed."

"I'm proud, very proud, that at Palantir we get arrows. Half the people attacking us, it'd be good if they spent two minutes on Wikipedia at least learning the talking points they're regurgitating."

video: @a16z

[a16z @a16z]

Technology only matters if it strengthens the people and institutions that hold society together.

In this conversation with a16z’s Katherine Boyle, Alex Karp discusses how AI is reshaping warfare, why Silicon Valley needs to take the stakes of this moment more seriously, and what it will take for America to remain both technologically dominant and socially cohesive.

@PalantirTech @KTmBoyle

[the VC almanac @theVCalmanac]

The difference between "similar" and "relevant" is why most enterprise RAG deployments quietly fail at scale.

[Nishkarsh @contextkingceo]

We've raised $6.5M to kill vector databases.

Every system today retrieves context the same way: vector search that stores everything as flat embeddings and returns whatever "feels" closest.

Similar, sure. Relevant? Almost never.

Embeddings can’t tell a Q3 renewal clause from a Q1 termination notice if the language is close enough.

A friend of mine asked his AI about a contract last week, and it returned a detailed, perfectly crafted answer pulled from a completely different client’s file.

Once you’re dealing with 10M+ documents, these mix-ups happen all the time.

VectorDB accuracy goes to shit.

We built @hydra_db for exactly this.

HydraDB builds an ontology-first context graph over your data, maps relationships between entities, understands the 'why' behind documents, and tracks how information evolves over time.

So when you ask about 'Apple,' it knows you mean the company you're serving as a customer. Not the fruit.

Even when a vector DB's similarity score says 0.94.

More below ⬇️

[Nishkarsh @contextkingceo]

We've raised $6.5M to kill vector databases.

Every system today retrieves context the same way: vector search that stores everything as flat embeddings and returns whatever "feels" closest.

Similar, sure. Relevant? Almost never.

Embeddings can’t tell a Q3 renewal clause from a Q1 termination notice if the language is close enough.

A friend of mine asked his AI about a contract last week, and it returned a detailed, perfectly crafted answer pulled from a completely different client’s file.

Once you’re dealing with 10M+ documents, these mix-ups happen all the time.

VectorDB accuracy goes to shit.

We built @hydra_db for exactly this.

HydraDB builds an ontology-first context graph over your data, maps relationships between entities, understands the 'why' behind documents, and tracks how information evolves over time.

So when you ask about 'Apple,' it knows you mean the company you're serving as a customer. Not the fruit.

Even when a vector DB's similarity score says 0.94.

More below ⬇️

[the VC almanac @theVCalmanac]

"Artificial intelligence would be the ultimate version of Google."

Larry Page in 2000.

"So if we had the ultimate search engine, it would understand everything on the web. It would understand exactly what you wanted. And it would give you the right thing."

"It would be able to answer any question, basically, because almost everything is on the web. So we're nowhere near doing that now. However, we can get incrementally closer to that. And that's basically what we work on."

"If you printed out our index, it would be 70 miles high now. We have about 6,000 computers. We have enough disk space to store 100 copies of the whole web."

"From an engineering, scientific standpoint, building things to make use of this is a really interesting intellectual exercise."

[Nic Cruz Patane @niccruzpatane]

Jim Cramer in 2010: “I’m not sure Tesla has a business plan that’s going to work, it’s not a smart investment”

Inverse Cramer is so real

[the VC almanac @theVCalmanac]

What is a VC?

"Most of these people we call VCs, they're not actually rich people with a lot of money that want to make some money. They're vehicles that raise money from institutional investors that have to return a certain amount of capital in a given amount of time."

"If I'm a professional VC and I have a $300 million fund, my mandate is to 2x or 3x that fund over the course of seven years."

"As an entrepreneur, it's very important to understand what the person on the other side of you is actually looking for. And what they're looking for is to return a fund of a certain size."

[the VC almanac @theVCalmanac]

Anduril’s @PalmerLuckey on AI and defense: It’s not up to unelected Silicon Valley executives to decide US policy.

“Don’t try to control policy by refusing to give the government the best tools.”

“You’ll see a lot of companies saying, ‘Here’s where we draw the line. This is what policy should be. This is what we’re going to allow the government to do and not do."

“That’s not the place of Silicon Valley companies. They shouldn’t be telling what their policy is.”

“Build the technology, advocate for policy.” “The US government and military make better decisions when they have better information.” “It’s not up to us.”

[the VC almanac @theVCalmanac]

Chamath: Work-life balance is the worst thing happened to young people.

"The first and most important thing is you have to be on Broadway."

"If you're into politics, you need to be in Washington D.C. If you want to be in finance, you need to get to New York or London. If you want to be in crypto, you probably need to be in Abu Dhabi. If you want to be in tech, you just need to be in Silicon Valley."

"There is no shortcut for any of these decisions. You have to be where the fish are."

"The number of young people I encounter who talk about all of these idiotic things like work-life balance. I don't even understand what that means."

Remote work is convenient. Being where it happens is how you win.

video: @chamath

[the VC almanac @theVCalmanac]

Palantir CEO Alex Karp on controversial uses of AI:

“Do you really think a warfighter is going to trust a software company that pulls the plug because something becomes controversial, with their life?”

“The small island of Silicon Valley that would love to decide what you eat, how you eat, and monetize all your data should not also decide who lives in a country and under what conditions.”

“The core issue is who decides?”