Scott

From the infinite potential of energy to the total actualization of entropy, intelligence charts a course for the pursuit of meaning, mission and love.

[Scott @scottstts]

Here it is, my recreation of the film Interstellar in @threejs, 13 iconic scenes, a cinematic montage

It’s live and open source, links below

90% of this was built with Codex, I used Opus 4.6 and Gemini 3.1 Pro for many prototypes of scenes, shaders, models (the ref/ dir)

[Scott @scottstts]

@TheGeorgePu I’m just surprised people still use GitHub copilot

[Scott @scottstts]

AI should be considered just like any other resources in your cost structure, it makes no sense to think “the more you spend the better”. This round of cost burnout is just businesses falling for and fueling the hype without realistic expectations and implementation strategies

Token subsidy is basically over, that forces companies to be smart about where to spend that token for real ROI, and it’s a good thing long term

[Business Insider @BusinessInsider]

Sam Altman said AI budgeting has recently become a "huge issue" for some companies, something that "never came up" earlier this year. https://t.co/P2zODBNmDp

[Scott @scottstts]

The self made billionaires have that aura, you don’t usually see that on inherited ones

[Chubby♨️ @kimmonismus]

How can this man have so much aura?

[Scott @scottstts]

I left codex running last night and went to sleep, find out it’s still running this morning, turns out limit reset🔥

[Tibo @thsottiaux]

Hi. Over the last 24 hours we had three separate small incidents that affected Codex reliability. Those are three too many and we are taking active steps for them to not reproduce.

I have reset usage limits for Codex across all paid plans. May the tokens flow again.

[Scott @scottstts]

Agents are still so brittle, it just easily makes a mistake right under your nose if you let it run wild

The main bottleneck about agent is still memory imo. When a person works on something, the memory is a constant influence factor filtering every output, like it’s baked in; agent memory is completely external and on demand, that just doesn’t work

I think the only way to truly solve agent memory is to do it at model layer not harness layer

[Scott @scottstts]

@gdb What does it mean? gpt 5.6? Super app? Any usage reset?

[Scott @scottstts]

Yeah I don’t have high hopes for Gemma 4 12B🫪

“Driving a dirty car just to go get it washed feels slightly contradictory. If you’re going to clean the vehicle, why not start by giving your legs a workout?”

[Scott @scottstts]

Just found out codex /goal run doesn’t go on forever after usage is out, after about 30 mins the agent is probably notified by system to wrap it up

And I do see a bug in this case where its summary doesn’t address the goal but the very first request in this session

[Scott @scottstts]

Gemma 4 12B dropped, pretty interesting, it’s a encoder free multimodal model. It turns input media into direct patchified vectors and down project to make them text token compatible (same hidden size), and the transformer backbone process them directly

[Google Gemma @googlegemma]

Meet Gemma 4 12B!

A unified, encoder-free multimodal model designed to bring high-performance intelligence directly to your laptop, and released under an Apache 2.0 license.

Bridging the gap between edge efficiency and advanced reasoning. Here is what’s new with Gemma 4 12B: 👇

[Scott @scottstts]

When I first watched this new ep of S9 rick & morty, I thought it was such a disappointment, then I saw this ep has 8.8 on IMDb, I read some comments and now I get it

There’s no cosmic adventure, no silly jokes, no mind bending twists, none of that classic r&m stuff

It’s just Rick after finishing the revenge that defined who he is completely becoming depressed. And even with a clean slate, he still defaulted back to who he was. He basically rediscovered himself from first principles, and that is a prison with no escape

This is something I myself can deeply relate, and so many others too

[Scott @scottstts]

HR invent problems that don’t exist, corporate America finally catching up

[Evan @StockMKTNewz]

$UBER just announced that it is cutting 23% of jobs in its "People" division that includes human resources, recruitment, workplace facilities and culture - Bloomberg

[Scott @scottstts]

I’m curious what 10x more productive means exactly?

Are you saying the amount of work for a whole year can be done in a month, and the other 11 months people just go on vacation?

Or is it that Bank of America will invent 9 more banks of the same size the same revenue employing the same amount of people?

Or the amount of work can be done 10x faster or with 10x less people, but they still appear to work all year or keep the same amount of roles that don’t actually do anything?

Or 10x just means “a bit more”?

Which is it?

[Scott @scottstts]

No you can’t be 10x more productive because you’re the bottleneck

[unusual_whales @unusual_whales]

Bank of America has said "you'll be 10x more productive with AI," per FORTUNE

[Scott @scottstts]

OpenAI makes /goal run uninterrupted even if you’re out of usage, it’s the only place where the old token utopia still partially exists

Generosity and thoughtfulness to users are rare these days

[Scott @scottstts]

@petergostev Fair trade

[Scott @scottstts]

With Opus 4.8 API pricing, assuming a 20:1 ratio of i/o tokens, and a small to medium task uses 2M input tokens and 100K output tokens which is pretty realistic

$1500 means 120 of these tasks a month at $12.5/task, that’s about 5 tasks/day

[Scott @scottstts]

Shocker

[Polymarket @Polymarket]

JUST IN: The EU’s €20 billion AI data center plan is reportedly facing delays and funding problems.

[Scott @scottstts]

Take one of your best and well polished products with decent complexity

Then a thought experiment: can you put an agent on a /goal loop to produce exactly this from scratch with 0 intervention, input, or reference?

Answer is no. Progress might be exponential, but the marginal difficulty of completing a task end to end is also exponentially higher, the last 0.0001% is infinitely more difficult than the first 90%

Together they just cancel out each other and nominal progress becomes roughly linear. If you add in societal frictions for adoption and diffusion, it easily becomes sub linear

[Scott @scottstts]

Well there you go. The AI washing layoffs laid bare. Overhire, mishire (I would also add “mal-hire”), and general mismanagement had been pervasive in these companies, no wonder the bloated orgs had so much bureaucratic friction that nothing gets done while OPEX skyrocketed

Everything you see here, they have nothing to do with AI disruption in any substantial way. It’s just decade long general org issues unwinding and self correcting

[John Loeber 🎢 @johnloeber]

Notes on 100+ Recent Technical Interviews

I interview a ton of engineers. Recruiting is the single most important technical CEO activity. Here are a bunch of impressions

1. There is a severe ZIRP engineering overhang that is currently washing out. They're getting laid off, managed out, etc. after having been massively overhired around 2020-2022. This is worst for Tier-2 big tech (think PayPal, Bill, etc.) but also FAANGs. These are overwhelmingly bad engineers.

2. This flood of unqualified but good-on-paper candidates makes this the hardest SF hiring market I have ever seen, due to the amount of nominally strong-looking candidates that you need to grind through.

3. I am highly skeptical of "AI as a cause for engineering layoffs". I think this is a large-scale polite fiction -- the companies don't want to admit they overhired, the engineers don't want to admit they are bad at their jobs. Everyone's blaming AI when it's really just the market rectifying itself.

4. Many of these engineers appear never to have had a real engineering function at their corporations. They're sitting in meetings, "making decisions about technology" but are unable to write software. I leave many interviews baffled by what exactly they were doing for so many years, let alone what their manager was doing.

5. I have interviewed some engineers from FAANG companies so shockingly nontechnical that I am forced to conclude that there is either (1) a lot of resume fraud going on or (2) that there are kickback grifts within those organizations -- people hiring their cousins and splitting the pay, that kind of thing. I have no other explanation.

6. There's a fun side-effect where after interviewing 20+ people from certain small but public companies, I actually feel like I am gaining a short sellers' advantage: there are financial technology companies out there that, knowing what I now know, I would never deposit a single dollar into.

8. Based on this "exhaust" data, and extrapolating a little bit, maybe aggressively so: I think folks like @pmarca are basically right when they say that ~every tech company is overstaffed by a factor of 2-4x. Whatever the reason -- staffing ahead of need, monopolizing certain engineer types (Google-style), headcount-driven promotion incentives, the reality is that a lot of these companies are not being run for the shareholders. The aggregate SBC expense is insane, and I expect this is going to get rectified eventually.

I'm sure that AI will play a role in rectifying this -- but I fear that people are going to blame AI for taking people's jobs when the reality is that the jobs were already long-gone, possibly always useless, but the highly-paid butts-in-seats remained. People will be mad at AI for taking away their lucrative sinecures. Maybe that's the same effect from a public policy perspective, but it feels different morally.

[Scott @scottstts]

@rajveerbach @scaling01 Just guessing based on this regression model, reaching upper end it might not be entirely accurate, the chart is from this paper

https://t.co/wmzugAQtmt