Kevin Patrick Mahaffey
@dropalltables
Followers
4K
Following
7K
Media
197
Statuses
4K
Truth-seeker. On a mission to drive cost of living to zero. Founder @ Lookout, Investor @ SNR (9x🦄). Believer in founding ideals of 🇺🇸
San Francisco, CA
Joined May 2008
The #1 question every founder and investor should be asking is: What’s defensible in AI? I'm publishing a series of posts that were previously only shared privately covering principles of defensibility in technology markets and how they apply to each layer of the AI stack. The
6
5
43
Can I sue the city of San Francisco for poisoning me with their virtue signaling while simultaneously failing to deliver on the basic services they’re funding?
New: San Francisco today will sue 10 of the biggest manufacturers of ultraprocessed food, arguing the companies knew their products were making people sick but sold them anyway. The fare makes up 70% of the country’s food supply. https://t.co/qNEQMVloDn
1
0
10
Make wine, not war. I would love to make competition between major nations the quality of our products, the breakthroughs of our research, the skill of our athletes, and the innovation of our industries. Nobody wins when power-hungry leaders bully their neighbors.
0
1
6
I present the “Husband Help Button”, vibed with @wabi: an application that gets me out of tricky matrimonial situations that may or may not occur between @calmastheocean and myself.
1
1
26
If "who has the best model" for any given AI use case changes on a week-to-week basis, all application layer users of AI must avoid model lock-in like the plague. With low model switching costs, foundation model defensibility must rely on other factors.
0
0
8
Sergey is a legend. How many would go full founder mode vs retire on their yacht?
Sergey Brin in founder mode actually saved Google. He had a big tiff inside Google, because Gemini wasn’t allowed to be used for coding. He told Sundar, “I can’t deal with these people. You have to deal with this.” Big companies always build bureaucracy. Sergey (and Larry)
1
0
10
The American Nuclear Renaissance is on.
0
1
6
Happy @sama resurrection day to all who celebrate.
0
0
2
Something I think people continue to have poor intuition for: The space of intelligences is large and animal intelligence (the only kind we've ever known) is only a single point, arising from a very specific kind of optimization that is fundamentally distinct from that of our
765
1K
12K
Friday night startup ideas that will immediately get me a $1B pre-pre-pre seed term sheet: 1. "Cursor for Cursers": Helps you create innovative ways to swear. 2. "Cursor for Curses": Avada Kedavra etc. 3. "Cursor for Bursars": Manages endowments. Wire instructions below:
3
1
14
Where are these high schools giving people As in pre-calc and calculus who can’t do middle-school level math? This feels like a failure of our K-12 education system more so than a failure of the university system.
the ucsd story was way worse than what i gleaned from the tweets, a whole public education pathway appears to have broken down because it's optimized around not having the appearance of inequity instead of teaching people things
1
0
5
This was always our motivation at Lookout: give defenders an economic advantage. Instead of human manual response processes, this goal forced us to automate as much as possible. What happened in cybersecurity can happen in defense.
$1000 *per round* to shoot drones with massive guns means the enemy sends ten $500 drones you can't ignore, then a few fast killers once you're out of ammo. Taking out drones should be *less expensive* than sending the drones. Inverting costs is how you win.
0
0
1
Gen 1 vs Gen 2 $3.5M BOM → $750K BOM. This platform will scale with Lumina for the next decade But the real unlock is that nearly 60% of its components (power distribution, hydraulic modules, cab, etc.) will carry across every future Lumina machine(excavators, etc).
The Gen-2 dozer exterior design is now officially locked in. It's more refined & emblematic of the Lumina brand. It’s full steam ahead as we put the vehicle into production and tooling early next year.
6
5
120
There are two mistakes people can make in evaluating AI’s impact on the economy. - direction: pessimists who point to problems with today’s technology (hallucination, cost, scaling) will likely be proven wrong with the steady march of research progress - duration: optimists who
"The share of companies citing quantifiable benefits from AI adoption has been steadily increasing based on our analysis." via @MorganStanley research
1
0
7