Something I think about a bunch is how @ycombinator has to pick startups (entrepreneurs with something of an idea) for each class. They, obviously, get some huge successes but also plenty of duds. And the duds obviously go the other direction too, rejects who would have been successes. This is just reality that there are limits on financial resource and, even more so, human capacity. But with AI driving down the time and cost to get to revenue - as @pmarca recently discussed the next gen of truly ai native startups - the number of ideas that will come to market should go parabolic, which should mean a bunch more successful startups to fuel the next 20-30 years of the US economy…if we don’t screw it up by anti-growth anti-ai anti-freedom restrictive regulations.
1
0
1
29
This is easy…they don’t want to pay the sky high price for DRAM. And there isn’t really a way to deliver profitable cloud services without memory density. The hyperscalers want to get through to 2029/2030 when enough new fab will be online to get memory prices down.
SITUATION DETECTED: Amazon CEO Andy Jassy says that AWS capacity for 2026 is completely sold out, 2027 is largely sold out, and "a good chunk" of 2028 is already sold as well.
0
0
0
37
I fundamentally believe that something that truly separates the winners and the average is the winners ability to understand the systems and games, and figure out to work the systems and play the games in advantageous ways - especially over the long term.
People who just reject playing usually end up as mids.
People who just reject playing usually end up as mids.
“I’m not doing XYZ bc it shouldn’t be that way despite it being beneficial for me” is a fucking epidemic in this country. The amount of people responding with that sentiment is depressing. Who gives a fuck if you don’t like how the financial system works; use it to your advantage
1
0
1
109
The S&P500 was up .41% and the NASDAQ was up over 1%…a whole weekend of doomer posts for not just nothing, but for serious losses if you took that bet on Friday afternoon or in the early overnight futures!
@ChadSlimeBased The market will close green on Monday.
0
0
0
41
Does anyone know the best way to move from a chat in Claude/ChatGPT to working in Claude Code/Codex. This should be transparent they are signed into the same account.
5
1
4
184
It’s took a couple guys a few months to build the Instagram MVP…so assuming you could now do it in a day or two, why haven’t we seen a bunch of Instagram like solutions emerge?
I’ll suggest it’s because you need to have a great idea and be genuinely passionate about that idea to the degree you are willing to go all in on it and drive it to greatness. And very few people have both the idea and the passion.
AI isn’t going to solve that. That’s a human thing. AI just means that Kevin Systrom and Mike Krieger could have gotten the code done a lot faster.
I’ll suggest it’s because you need to have a great idea and be genuinely passionate about that idea to the degree you are willing to go all in on it and drive it to greatness. And very few people have both the idea and the passion.
AI isn’t going to solve that. That’s a human thing. AI just means that Kevin Systrom and Mike Krieger could have gotten the code done a lot faster.
0
0
0
34
Nicolas Nowinski retweeted
I’ve been to the Louvre. I’ve stood in front of the Winged Victory of Samothrace. But my children have not.
Europe, France, men of the West, are you going to protect the inheritance for future generations? Or let the plunderers, for whom you threw open the gates, continue to destroy it?
Europe, France, men of the West, are you going to protect the inheritance for future generations? Or let the plunderers, for whom you threw open the gates, continue to destroy it?
173
605
4,765
123,991
Why is this an interview question in 2026? You can just ask your agent and get the answer.
NETWORKING INTERVIEW QUESTION
Your computer has an IP address of 192.168.1.10/24.
You try to communicate with 192.168.2.20.
What does your computer need to reach that device?
A) A DNS server
B) A router
C) A network switch
D) A DHCP server
What's your answer: A, B, C, or D?
Easy to memorize. Harder to explain.
2
0
0
145
@nicknow @yourclouddude Your PC is on 192.168.1.0/24. The target 192.168.2.20 is on a different subnet, so traffic cannot stay local. A router (default gateway) is required to forward packets between networks. Switch only works inside the same subnet; DNS and DHCP are unrelated to routing. Answer: B.
0
0
0
13
Nicolas Nowinski retweeted
Hi, I'm Felix. I'm an eng lead working on Claude. I love watching people take flight and do things they never thought possible.
We often describe that as "being more ambitious" when using AI, which is tricky advice to follow. I wrote a tutorial:
felixrieseberg.com/how-to-be-more…
We often describe that as "being more ambitious" when using AI, which is tricky advice to follow. I wrote a tutorial:
felixrieseberg.com/how-to-be-more…
405
2,539
24,237
5,377,749
Nicolas Nowinski retweeted
I know tech folk have only the haziest understanding of the real world, but no S&P500 co is throwing out Salesforce to replace it with a CRM some random dude vibe coded in his bedroom.
If you don't understand why, start hanging out with much better marketers...
If you don't understand why, start hanging out with much better marketers...
Saas companies died the same week that globally we all started 3d printing our clothes and furniture.
It’s never going to happen
23
9
179
7,354
Nicolas Nowinski retweeted
Software is dead in that same way restaurants are dead because i can grow my own vegetables and own an air fryer
98
285
4,431
230,619
This is a really good way of thinking about what we are seeing in software dev careers today.
@nateberkopec I know why but maybe I am delusional.
People operate in two ways. Mindset or rubric. I work with some people who need to be told what to do or given a workflow to do their job. These are software engineers. Then I have people who I can give a mindset to and they can approach new and novel problems and solve them.
I think what you're describing is the former. People who only solve surface-level problems because they ticked a checkbox.
0
0
3
66
“Try rebooting it” is now “ask an agent.”
Why? Because rebooting only works sometimes and when it does you never learn why it happened and how to prevent it from happening again.
But the agent usually can get to the root cause and help prevent it in the future.
Why? Because rebooting only works sometimes and when it does you never learn why it happened and how to prevent it from happening again.
But the agent usually can get to the root cause and help prevent it in the future.
0
0
0
37
Nicolas Nowinski retweeted
I'd expand the moats left to:
- Proprietary data or content: data nobody else has, or unique content you can create
- Network effects / community: users make it more valuable for each other
- Audience / distribution: you can get attention on demand
- Physical component: hardware, locations, real-world operations
- Capital: owning assets and investing (esp into the AI companies as they are and will suck up large parts of the economy)
- Regulatory: licenses and legal barriers
- Proprietary data or content: data nobody else has, or unique content you can create
- Network effects / community: users make it more valuable for each other
- Audience / distribution: you can get attention on demand
- Physical component: hardware, locations, real-world operations
- Capital: owning assets and investing (esp into the AI companies as they are and will suck up large parts of the economy)
- Regulatory: licenses and legal barriers
Technical moats are mostly over at this point. If you’re building something that is meant to be durable, the only moats now are:
• Proprietary data
• Network effects
• A physical component
293
323
4,274
629,026
I think this may be entirely true and also will not matter. As AI becomes more ubiquitous many of these people will change their views.
Worse, for them though, is that the early adopters will be financially benefiting from compounding for decades to come.
Worse, for them though, is that the early adopters will be financially benefiting from compounding for decades to come.
You are underestimating how much people out there hate AI
I was at a wedding this weekend and the consensus take among people I met was that using AI for anything is morally wrong and generally inexcusable
Guess I’m living in a bubble bc I was shocked…?
0
0
0
79
Everyone who is saying all the SaaS firms will be replaced with custom software seem to have forgotten the last S is for Service. A huge reason SaaS took off wasn’t that it was an amazingly great advance in functionality…it’s that you didn’t have to install it and then own it.
0
0
1
36
1) This is almost inevitably true.
2) We are in the early days of RSI where labs will be see the biggest benefit, that benefit will rapidly shrink over time.
3) RSI won’t lead to some type uber super intelligence taking over the world.
So, bottom line, we’ll get better models and that’s nice…but that’s about it.
2) We are in the early days of RSI where labs will be see the biggest benefit, that benefit will rapidly shrink over time.
3) RSI won’t lead to some type uber super intelligence taking over the world.
So, bottom line, we’ll get better models and that’s nice…but that’s about it.
Let's take this seriously: even Google employees are now clearly stating that it seems to be an industry consensus that all Frontier Labs are managing "RSI loops." In other words, recursive self-improvement is being implemented by everyone.
Furthermore, Fable 5.5 will be the best model for Anthropic's IPO, but internally, OpenAI's "Bel" will undoubtedly overtake Anthropic by the end of the year.
RSI is here, and internally, the labs are miles ahead.
0
0
0
53
Something that gets missed is how much software is being written today and never seen outside. Open source is exploding - heck I just pushed my fourth Omarchy plugin to GitHub - but the bigger deal is how much is really bespoke for personal/single org - use case.
1
0
3
87
Most of it would never have been built before. If no SaaS/COTS solution existed it would be handled manually (Excel/Sheets, Email, File Shares, etc.)
Building publicly released software is still real work, but personal/internal is incredibly quick.
Building publicly released software is still real work, but personal/internal is incredibly quick.
1
0
0
18