Any founder or investor should read Sam Altman’s post reflecting on the past year at OpenAI. While the headline is around Superintelligence 👇🏼, what I found most

interesting were the nuggets of wisdom on building and running a startup in a market that is changing so fast.

It always nice to remind ourselves that no matter how much great content you read on X or Linkedin or wherever or no matter who advises you, it’s hard to know what’s really going on unless you are living it day to day. Sure, we can learn from the greats in the past, but remember to be an artist, take the best from others, remix and come up with your own way of building things or running a company. One day other founders will say I want to build like X, but you know deep down in your ❤️, that they should definitely learn but only take the pieces that work for them. Joao from CrewAI, who is super early in his journey, had this great post the other day.

Keep experimenting, keep trying new things, don’t anchor to the past, look to the future!

Speaking of the future…no one is seeing or creating it better than Jensen Huang from Nvidia. See his super clear thesis laid out below from Generative AI to Agentic AI to Physical AI. What’s even more interesting is how the other great CEOs are lavishing praise on Jensen’s must watch keynote here.

Never content to be a GPU creator, here’s how Jensen is going deeper into the software layer as he proclaims our agentic future is here, one that is poised to be a multi-trillion-dollar industry 🤯. This is the very same future that Sam believes in as he stated that we may see the “first AI agents join the workforce in 2025.”

And I’m 🔥 up that CrewAI is a huge part of this announcement.

I remember when Joao and team had some of their first Zooms with the Nvidia team to make sure they could build the right use cases fast enough to be part of this launch - huge congrats on spending those many sleepless nights making this happen! If interested, here’s more on what and how Nvidia will work with agentic frameworks like CrewAI.

And here’s one of the demos CrewAI put together with Nvidia showcasing how 2 crews of agents can help engineers write docs!

As always, 🙏🏼 for reading and please share with your friends and colleagues!

Scaling Startups

#my annual reminder…

#lean startup vibes heading into 2025

#YC growth hacks - how to do cold email the right way

#aligns with my post from last week, especially #3 - GPs using secondaries to generate DPI

read #6 from last week

What’s 🔥 in Enterprise IT/VC #427

Ed Sim · Jan 4

Welcome to 2025 and the death of venture capital as we know it 🤣. Death is an extreme, but we can all agree that it must adapt. There was lots of discussion online this week about the future of VC and exits, so here’s my take.

Read full story

#I highlighted this last week as well but only shared #1 - Excellence is the capacity to take pain - here’s the full list of the 15 recurring traits of history’s greatest founders - one of best podcasts I’ve listened to in awhile!

#early is best place to be especially first for any winner, why the “race to be first” and get pole position on cap table in full swing esp. with the multistage funds viewing “Inception” rounds as options (more from What’s 🔥 #365)

#Jensen Huang puts it best, “you want a company that’s as large as necessary to do the job well, but to be as small as possible,” and not bogged down by over management and processes.” (from the Nvidia Way by Tae Kim)

Enterprise Tech

#💯 this portability is a huge unlock and check out what dylib.so (a port co) is doing with MCP.run to make these even more

Here’s the excerpt to save you the time:

MCP is an open protocol that standardizes how applications provide context to LLMs. Think of MCP like a USB-C port for AI applications. Just as USB-C provides a standardized way to connect your devices to various peripherals and accessories, MCP provides a standardized way to connect AI models to different data sources and tools.

Why MCP?
MCP helps you build agents and complex workflows on top of LLMs. LLMs frequently need to integrate with data and tools, and MCP provides:

* A growing list of pre-built integrations that your LLM can directly plug into
* The flexibility to switch between LLM providers and vendors
* Best practices for securing your data within your infrastructure

#if folks get behind this, this will be big…here’s the future roadmap

#second order effects in action - what happens when more AI generated code? more code that can F things up, more code to maintain and more code to review!

#great 🧵 from a seasoned multi-time founder of enterprise startups

#amazing how fast the founder of Rippling, Parker Conrad, jumped on this tweet as company is worth $13.5B - worth a read, no company is perfect, but sounds like some users have some legitimate gripes - will be interesting to see how this evolves 👇🏼 - main issue is speed! users want fast experiences

#reminder - CIOs/CISOs don’t get paid to take on technical risk while startups should be building for the future - listen in on this 2 minute clip

#agreed, one of my favorite tools also

#What are the many ways CISOs think about AI and security? Read more from Elevance Health head of security Sekhar Nagasundaram (Fortune 20) in FastForward

The company recognizes that attackers will be using AI, but it is also looking at AI as a way to scale and get things done faster. That means using Generative AI to understand the nature of threats and attacks and get to the root of a problem faster. Instead of a long laborious search, they can simply interact with the system by asking questions and get answers much more quickly.

The goal is to move from a static or even adaptive SOC (security operations center) to a cognitive one, but he says that’s a work in progress.

#who’s building the infra to manage and auth these? Sound familiar? Yes, second order effects I keep writing about on what happens when…

#👀 as the above happens

#Quantum bubble? Great post on what is quantum and timing…

The misunderstanding seems to be the idea that quantum will ”change everything”. No. It will help a few nerdy people, like me, study cryptography and odd, but useless mathematical problems, more easily. That is not a big industry

Markets

#Crowdstrike fully recovers from the outage - $30B of market cap back, what doesn’t kill you makes you stronger (FT)

CrowdStrike, the cyber security company behind the botched update that took down millions of Windows PCs and servers in July, has more than recovered the $30bn in market value it shed in the wake of the crisis.

The company’s shares, which plunged by more than a third in the two weeks after the incident, are now worth more than they were on the day before what US House Homeland Security Committee chair Mark Green called the “largest IT outage in history”.

CrowdStrike chief executive George Kurtz told the Financial Times the company had bounced back by turning the crisis into “a competitive advantage”. 

He said the incident — which stranded airline passengers, interrupted hospital appointments and took broadcasters off air around the globe — had not dented customers’ trust.

“Customers are staying with us,” Kurtz said in an interview after the company’s latest earnings report. “We had one customer say that broken bones heal stronger and they don’t expect this to reoccur. Conversely, from a competitor standpoint, that hasn’t gone through something like this, there’s probably more risk.”

#👀 terrible to see

While they are relatively rare in US venture investing, more than 80 per cent of venture and private equity deals in China contain redemption provisions, according to Shanghai-based law firm Lifeng Partners estimates.