Zlatan retweeted
The longer I immerse myself in robotics, the more I feel deeply in heart the urge to become full-stack. Robotics is a system problem, and pure model training counts less than even 20% of what really matters.
A good working system relies on so many things to be correct: how and what data you collect, what observation/action you choose to use, how you model your robot/environment, what hardware assumption/limits you have, what objective you train on, how to deal with latency, just to name a few. And often times, the bottleneck dynamically shifts, so you have to be sharp enough to pick the right problem to tackle and solve at the right time to push overall progress forward. Remember, one part fails and the entire thing won’t work.
You need to have an architectural and high-level view to nudge around for success. This is what makes robotics so hard, but same time so fascinating.
Zlatan retweeted
Made myself a little pedal prototyping box 🎸
Breadboard on top, and all the knobs, switches, footswitch and jacks are already wired to a header next to it. So trying a new circuit is just a few jumper wires, no soldering
Free to print on MakerWorld 👇
makerworld.com/en/models/364…
Zlatan retweeted
Me: I want to show this app transition on a marketing site. Should I do a screen recording?
Claude: Nah man. Let me port the Metal shader to WebGL as instanced quads, rewrite your shuffle algorithm in JS, write a Swift tool that bakes 132 covers into a sprite sheet, encode it as AVIF, benchmark two encoders against each other with SSIM, fake trackpad momentum with exponential decay, and dim the neighbouring covers so you don’t notice they’re blurry. 5 mins max.
Me: Cool, cool…
Naivety and AI allows people to approach challenges with creativity and optimism helping them achieve much greater things. Experience is almost a negative as it holds back people from trying things because they conclude it doesn't work given their experience.
James Maynard (Fields Medal 2022) says that he was shocked by the proof of the quasi-Riemann hypothesis (from the OpenAI math dump) because he held a very strong prior belief that the problem required the development of an entirely new theory.
He felt the solution was completely out of reach for current mathematical techniques, so seeing it included in the results drop was a massive shock to him. He explains that the proof turned out to rely on the theory of metaplectic forms, which had been developed in the latter half of the 20th century but was an underappreciated and slightly obscure part of the field that most mathematicians had not realized was powerful enough to resolve the hypothesis.
Source: Oxford Mathematics in Conversation on AI - Ben Green and James Maynard with Jan Obloj uploaded on Youtube 6 days ago.
73
90% of frontend engineers were slack anyway. Not technical enough to understand the full stack nor had taste to understand quality of the interface. The 10% remaining that can do either or both will be the most valuable engineers you can hire
Engineering is changing: FDE is ascendant while mobile and front-end roles are disappearing.
Open mobile and front-end engineers jobs have been nearly halved in the past year!
77
Zlatan retweeted
I fucking love working with engineers man.
While a lot of stuff in engineering is a judgement call, a lot of it isn’t.
And I just love that engineering practice and culture never hides that.
Literally every other vocation has some element of face saving behavior when proven wrong in public.
And so many good engineering cultures are like
> the answer is 100
> no actually it’s 200
> oh you’re right I’m wrong
And then people go on with the right information.
It’s truly insane how common in some vocations someone can call the sky green in a meeting and either nobody will correct them on it or they won’t admit it and “take it offline and circle back” without ever admitting fault.
And the biggest thing is that so many engineering leaders will get corrected by subordinates.
That’s really fucking rare out there.
In other vocations corrections so often go strictly down the org chart.
Lots of companies have it all and get everything right. Engineering lead companies often do.
But so many people out only complain about “engineer” or “engineering brain” and it’s literally just people who won’t let you say random false shit constantly.
Parkinson's Law, AI edition: work expands to fill the space available. Make everyone 10x faster and you don't get 10x fewer people. You get 10x more work.
Just spent some time catching up with an old coworker, who now works at the VP level at a boutique investment bank
Was asking him about analyst use of AI and thought his insight was quite fascinating
More people are using AI now vs just few months ago, and it absolutely cuts down a lot of time from mundane tasks
Just tedious things like arranging logos on a deck or managing a contact log for investors would require analyst resources behind it.
Now an Associate or VP with Claude can just do it by themselves with minimal effort
However, the work hours have not gone down at all for the analysts. Since senior folks know about the capabilities of AI, they are now staffing analysts across more deals at any given time
Not only that, MDs are chasing after new mandates more often than before.
Directors are coming up with new tasks for their analysts, like maintaining tear sheets or comp sets on niche verticals. VPs are asking for more pages and trying to show more analysis to clients than before
In summary, the duration of any single task has gone down because of AI.
However, this has been offset with new work that the firm and group simply never had capacity to do before
Instead of reducing hours, it has opened up room for his group to chase after more revenue and provide better/faster work to existing clients.
Suspect this is also what is happening across many industries and companies, not just in banking or finance
Walked away with the reinforced belief that the entire job loss narrative of AI is just doomerism
95
Zlatan retweeted
This sounds extreme and I wouldn’t have believed it even a year ago, but we’re starting to see plenty of examples of productivity gains that are an order of magnitude or two above output levels pre-AI.
With the right harness, a deep understanding of your domain of what you want to build, likely a relatively new codebase (with some exceptions), a fluency of how to work with multiple agents at once, and a setup that makes it so you can test and deploy with limited friction, outputs on this level feel realistic in coding.
What’s crazy is this gap doesn’t appear to be diminishing over time. You’d think that we’d see a convergence in the lines because of AI, but it turns out the more depth you had before -and willingness now to take a different approach- gives you crazy leverage. And that is only compounding as the models can take on larger work and harder problems.
Right now this is most pronounced in software engineering for the obvious reasons. But there’s no reason we don’t start to see this in a number of other computer-use-sensitive domains like other scientific fields, AEC, finance, cyber, and more. Going to get quite wild.
Growing up speaking multiple languages I have this test for when I’ve adapted to a new location and that is thinking and dreaming in that language.
Adapting to AI as a software engineer feels similar. You stop thinking and visualizing lines of code but only think in systems, entities and how they interact with each other.
1
4
158
Zlatan retweeted
narrative violation: preferences will swing back from DIY vibe-coded software to super secure industry-leading SaaS solutions.
while it's super easy to build anything, endlessly protecting and securing it is super hard...and teams want to sleep at night.
Zlatan retweeted
A moat never made a castle impregnable. It slowed attackers down. The wider and deeper it was, the more time it bought.
The industry made it sound like “moat” means high immunity from competition.
But I think moats will remain as long as people are involved.
But also means properly maintained software is safer than ever
3% of “critical” software security bugs, in all of recorded history, have been found in the last month.
This is on pace to find another ~100,000 bugs this year alone, adding ~26% more vulnerabilities to the all time total
Do you realize how fucking insane that is?
100
Zlatan retweeted
3% of “critical” software security bugs, in all of recorded history, have been found in the last month.
This is on pace to find another ~100,000 bugs this year alone, adding ~26% more vulnerabilities to the all time total
Do you realize how fucking insane that is?
Even if rental vacancy rates jump to +9% not much will happen as long as you have a +10y outlook
Dinner with an exec/friend at a large national property management company tonight. This person oversees ~150k apartments; access to an insane amount of data.
Current multifamily market in most of the country is worse than S&L, .com bust, or the GFC. Worst they’ve seen.
Class A mostly fine, decent rent growth. However:
- credit scores at class A have gone down meaningfully over the past 3 years.
- more people are paying rent via credit card.
- average bank account balance noticeably lower than even 2 years ago
- class A tenants stretched thin, living beyond their means. 20k comes in, 20k goes out every month.
- starting to see these renters move down a tier
Class B:
- simply too many apartments, we built too much new stuff. Ripple effect is enormous.
- fewer renters tomorrow than today. Same store foot traffic noticeably down year over year. Some markets down 40-50%
- considering dropping rents below $700 a month on some 1beds in major southern markets.
Class C:
- looking at degrading NOI. Rents aren’t growing for a long time, but expenses keep creeping up
- more renters are one missed paycheck away from disaster than they can remember seeing in decades
- tenants aren’t paying rent because they simply don’t have it
- wage growth has not kept up
Lenders:
- some major investment funds, household names, haven’t paid their mortgages in 2-3 years. Banks won’t foreclose because “X is a big client.” Mexican standoff.
- a roof leak gets ignored, becomes water damage, then mold, then a down building. Buildings are rapidly deteriorating as some owners choose cash preservation over funding losses.
- number of pre-foreclosure and REO deals is the highest as a % of total units they can recall.
- data providers know it’s bad, but they’re paid to sell optimism to owners/lenders. So every forecast is “it will probably get a little better next year maybe” and that gets regurgitated at investment committees all over.
I asked “how do you think this ends?”
“Not sure. My gut says slowly then suddenly, and things will get really bad. Been doing this 40 years, you can feel it in the air. But, there’s a chance it’s just a decade of tolerable pain. I’m praying for that.”
1
55





























