Related posts
Peter Steinberger Blog
The the words it uses that this site has seen least often elsewhere. Posts are matched on those words alone — nothing here is a summary of this one.
Top people
Showing Profile →
Hiding
Hiding
20 September
19 September
18 September
17 September
-
korrents.com
As AI inference costs fall, rising customer expectations will keep service margins from converging toward 100%.Their words
Margins will not climb to 100%. As inference gets cheaper, buyers will ask more of their agents, shifting the equilibrium over time in response to competition.
16 September
15 September
From one piece @paulg on X 2 beliefs · x.com
-
Their words
Someone needs to define the unit, perhaps using a chain of increasingly hard problems, each pair of which can be solved by a single model.
-
korrents.com
Tokens are not the true unit of AI inference, because better models yield more problem-solving per token.Their words
Although you pay for AI by the token, that's not the unit of inference, because you get more problem solving per token as models improve.
14 September
13 September
12 September
11 September
10 September
9 September
8 September
4 September
3 September
-
Their words
there's still a lot of startups in this batch that are not shipping fast enough. And so obviously there's variation in shipping speed. It's not just the rate at which you can produce things. You have to think of these ideas first, right?
-
Their words
It’s not orderly. But neither is water flowing in a stream. It flows around obstacles. Everything is opportunistic, selfishly taking the effective path, exploiting available gaps. It’s natural.
-
open sourceAnthropicOpenAIChina
Their words
in this whole USA vs China thing OpenAI and Anthropic aren't relevant because they're positioned differently them building better models doesn't hurt china at all the competitor has to be - american - open source - enough compute to do inference at scale that can shift things
1 September
-
korrents.com
Stoke's reusable launch vehicles will be shipping Starcloud's data centers into orbit within a few yearsTheir words
In a few years, Stoke's reusable launch vehicles are going to be shipping Starcloud's data centers into orbit.
30 August
-
Their words
While Western export bans have severed Russia from much of the auto market, they have not tempered the demand for brand-name SUVs and trucks. That has given rise to a sophisticated network of criminals that steals cars, hides them in shipping containers and sends them to Russia, often by way of the Middle East
26 August
From one piece Why performant code matters (but gets widely ignored), with Casey Muratori 2 beliefs, in the piece's order there
-
Their words
They've got blog posts of we had to rewrite this whole thing because the performance was bad. If it was always hotspots that made your performance bad, you'd never have to rewrite the whole thing. So, we know that that doesn't work anymore.
-
Their words
It's it's so massive that there is no way that your game will be organically noticed anymore pretty much period.
25 August
23 August
-
korrents.com
Demand for intelligence is highly elastic — every fall in inference cost is met with rapidly growing usage.Their words
the demand for intelligence is highly elastic: as inference costs fall, usage grows rapidly.
20 August
14 August
-
Their words
buuuut my roadmap is’nt ten times shorter. and I’m definitely not ten times better at deciding what is worth building. teams haven’t started casually shipping a year of product work every month. you can look around and and it’s hard to say what software has gotten meaningfully better in the last year.
12 August
10 August
2 August
-
Their words
there are people who are really big on road maps and there are people who are really big on iteration. I I think the honest answer is you've got to do both. You can't over-index on either.
30 July
From one piece Jeff Dean: The 1% Rule for Building in AI 3 beliefs, in the piece's order there
-
Their words
And um if you build a specialized chip for low precision dense linear algebra and can't do anything else that turns out to be really useful for machine learning inference uh even though it can't run Chrome or Word or whatever.
-
Their words
Um, and that's a very very useful general technique is you know inference time compute to perform search over plausible ways of solving the problem that can get much much higher performance or much more reliability in longunning agent flows.
-
Their words
Yeah, I mean it's a little different, but I think uh you're going to see more and more uh uh high performance and um low energy uh inference hardware systems because I think everyone is now realizing that inference is the key to making you know these agent-based systems be available to more and more people and that latency is really important and that specialization of the hardware is a really key way you can make uh things that are more energy efficient and lower latency than more general purpose uh computational devices like say GPUs or TPUs
29 July
-
Recommendstheir ownrcmnd.app
SuperlogicalTheir words
Superlogical will begin by shipping a terminal multiplexer. I'm pouring my years of experience building a terminal and studying the potential and limitations of other multiplexers into something new, powerful, and, of course, fast.
22 July
21 July
19 July
From one piece Why Netflix is betting on systems thinkers—not specialists—in the AI era | Elizabeth Stone (CPTO) 2 beliefs, in the piece's order there
-
korrents.com
That AI lets anyone prototype does not mean everyone should be shipping code to production.Their words
Do I believe that means anyone should be shipping code to production? That everyone should actually be doing everything? Probably not. But I think that it's good for people to be exploring what's possible.
-
Their words
I get really nervous about having different design languages or different types of user interactions and shipping Frankensteins, basically. So, designers need to then be the people we're hiring again for design systems thinking.
18 July
-
Their words
a smaller model at a higher reasoning effort can sometimes reach a similar score as a larger model at a lower reasoning effort
Controlling Reasoning Effort in LLMsmagazine.sebastianraschka.com
15 July
From one piece Context engineering with Dex Horthy 3 beliefs, in the piece's order there
-
Their words
And so it's like I mean getting into Eli Goldrat and the goal is like optimizing for utilization and efficiency of one node in your factory rather than the end to end goal of like how do we ship value and things that people like that are stable and like will last a long time. But that's my idea of token harder
-
Their words
You'll notice what I said was not use loops to ship the features that users want. We use loops to actually improve the codebase quality and we read all the code because we care about how it's architected and we care not just about the system architecture but what I would call the program design
-
Their words
We tried this. We built a lights off software factory in July of 2025 and by November we had shut it down. I think it takes about three to six months of you shipping all the time with nobody reading the code before you realize like, wow, this is getting way worse and it's easier to start over than it is to fix it.
30 June
-
Their words
It becomes insufferable like as a mathematician because you you would basically be like I'm every single time I see one of these I kind of don't know if it's worth my time even if 99 out of 100 of them are right.
24 June
From one piece Tech interviews with NeetCode 2 beliefs, in the piece's order there
-
Their words
If you're going so fast, you're not measuring the impact of the changes that you're making, you don't have time to do that cuz you're just focused on shipping and then things regress and things get worse and we've seen that at Anthropic recently the last like month or maybe more than that where things have regressed.
-
Their words
But in terms of like features, like a website like you can just throw features in there nowadays that nobody really cares about and you can you can do it so quickly. Like a new feature every single day, but do people actually care about that? Is that making it better? It could be making things worse.
15 June
14 June
-
Their words
there's a difference between belief and hope. Hope is confidence without basis. Hope is just, you know, it's it's a prayer, but it's but it's it's not it's not founded in anything that your your lived experience.
9 June
-
Their words
And then for those of you who are skeptical about international organizations, here's one you've never heard of. The International Organization for Standardization then standardized these container sizes to fit one if by truck, two if by railway cars, and thousands if by sea. And this just plummets transport costs.
7 June
-
Their words
And if you're doing a 1.0 and the world hasn't seen you, you're not going to get that from consumers, ever. You have to ship it, and you have to build the entire kind of ecosystem so those consumers see it in the fullness so that when they do the evaluation and they spend their own money, then you're getting real feedback.
4 June
3 June
-
Their words
I think the number one success criteria was docker so remember there was mos and msosphere and they had their own runtime corp had come out with nomad and they had their own runtime but the biggest runtime that had already got global consensus was docker
27 May
From one piece Building OpenCode with Dax Raad 4 beliefs, in the piece's order there
-
Their words
The moment you ship something, you're stuck supporting it forever. And by supporting, it means any future feature you build is going to like interact with it. So you still have to be very conservative with what you put out there. It's hard to undo anything. Just cuz we can ship 10 times more doesn't mean we have 10 times as many good ideas to ship out there.
-
Their words
cuz because we rent GPUs at scale to run the models and we still use middleman by the way. So we're not like going all the way down to the down to the floor. Even for us there are some models the sticker price and the cost to us there's like an 80% margin in there.
-
Their words
There's always negative sentiment that exists for any business that's getting hyped. They have no incentive to correct it. Um so again it's complicated because I know the training costs are a big part of it. Uh the R&D department is is hugely expensive but long-term inference makes sense as a business and I think it it always will.
+ 1 more
-
Their words
The demand for inference is growing. So, like I don't think it's linearly growing. I think it might even be exponentially growing. But we haven't made our production of GPUs grow exponentially. That's like kind of a linear process. So as those lines intersect, there's going to be uh tightening.
-
Dislikedaffiliate linkrcmnd.app
Synergy Alpha Elite dumbbell rackTheir words
Summary: Shipping damage. Residue on some of the components. Loud pings and pops from the assembled weight rack. A design that looked great and saved space but was painfully unusable. Restocking fee. Non-refunded shipping fees. Ghosted me about the refund.
25 May
24 May
-
Their words
But the coding models have gotten good enough that he can pair the kind of the technical knowledge he does have with his really spiky product sense and sense for writing and sense for users.
22 May
20 May
-
Their words
My opinion is that the least mature area is front end. There have been some attempts to compile Rust to Web Assembly and then run it on the web as a front end as a replacement for TypeScript. But if I was writing a web server, I would totally use Rust for the back end and TypeScript for the front end. I would not really go the web assembly route.
13 May
-
Their words
because if you were to force AI to write a type annotation on everything, then it would probably get it wrong more often because now it has to keep track of all these types and and it and it has to just repeat itself over and over and over, right? And so, types are important where there's no context.
6 May
24 April
15 April
-
Their words
Pick your ASIC team where you can say I can bet the farm of I can bet my entire business that you will be here for me every single year. Your cost, your token cost will decrease by an order of magnitude every single year. I can count on it like I can count on the clock.
6 April
-
Recommendsrcmnd.app
Dell XPS (Panther Lake)Their words
Give the new Dell XPS series, or any of the other laptops shipping with Panther Lake, a try. I think you'll be as impressed as I've been.
29 March
23 March
-
Their words
that was always illogical to me because inference is thinking, and I think thinking is hard. Thinking is way harder than reading.
13 March
From one piece Dylan Patel — The single biggest bottleneck to scaling AI compute 3 beliefs, in the piece's order there
-
Their words
Um, just shipping out all the engineers and blowing up the fabs means China has a stronger semiconductor supply chain than the rest of the world, right?
-
Their words
They could release claw slow mode and have an increase in tokens per dollar by a significant amount. Um they could probably like reduce the price of Opus 46 by you know 4x 5x and reduce the speed by another by maybe just like 2x like the curve on inference throughput versus speed is there already just on hm um and yet they don't um because no one actually wants to use a slow model
-
Their words
So when you look at inference at let's say 100 tokens a second for deepseek and kimk 2.5 hopper versus blackwell the performance difference is on the order of 20x
12 March
11 March
-
Their words
The best feature we can add for the player is shipping.
13 February
-
korrents.com
Longer context is an engineering and inference problem, not a research problem — nothing prevents it from working.Their words
There's There's nothing preventing longer context from working. You just have to train at longer context and then learn to to serve them at inference. And both of those are engineering problems that we are working on and that I would assume others are working on as well.
10 February
-
Their words
The economics of orbital “datacenters” or essentially glorified Starlink satellites with a bunch of GPUs attached are likely to be even better than Starlink.
Space AI: I guess we’re doing Moon factories nowcaseyhandmer.wordpress.com
24 January
21 January
-
Usesrcmnd.app
WaterpikTheir words
Waterpik: the first thing I do when waking up is use a water pick. It uses a high-pressure stream of water to remove food, plaque, and bacteria from hard-to-reach places. Morning & night.
18 January
30 December 2025
23 December 2025
8 December 2025
-
Their words
By 2045, natural gas will be used as LNG primarily for high performance supersonic aviation, shipping, and industrial heat.
4 December 2025
8 October 2025
-
Usesrcmnd.app
Elgato Stream Deck+Their words
Used for streaming and daily automation tasks.
17 September 2025
15 September 2025
1 September 2025
-
korrents.com
Shipping a half-baked product permanently damages customer trust, even after it is later fixed.Their words
When AWS ships a service which is half-baked, it diminishes customer trust in AWS as a whole; even if the problems in that service ultimately get corrected (either by fixing them or in some cases by simply getting rid of a service which should never have existed in the first place) the memory of a failed launch will live on in customers' minds for years to come.
24 August 2025
15 August 2025
-
Their words
actually it turns out that you can significantly decrease power consumption with a very small reduction in overall compute. So if you if you've got like three really bad days in a row or something, you can actually just like you can dial back your power usage quite a lot without compromising your inference or or um or training.
25 June 2025
4 June 2025
30 April 2025
-
korrents.com
Major Linux distributions are increasingly shipping system libraries with frame pointers enabled by defaultTheir words
This is getting easier in the lastest versions of Fedora and Ubuntu (e.g., Ubuntu 24.04 LTS) which are shipping system libraries with frame pointers by default.
23 March 2025
-
Their words
But this is the hard part. When you're when you're doing that market widening, you're not solution deepening. And so your perceived product velocity may decrease. You can avoid some of these things with some smart technology decisions, but mostly you just have to grind through it and it is worth it to get to the other side.
3 February 2025
-
Their words
OpenAI has a fantastic margin. When they're doing inference, their gross margins are north of 75%. So that's a four to five X factor right there of the cost difference, is that OpenAI is just making crazy amounts of money because they're the only one with the capability.
30 January 2025
-
Likedrcmnd.app
MacBook Pro M1 MaxTheir words
This allows me to run models locally on my MacBook Pro M1 Max. With the 64GB of RAM it has, it’s a pretty potent machine for basic inference despite it being three years old.
11 January 2025
31 December 2024
-
Recommendsrcmnd.app
Is AI progress slowing down?Their words
To understand more about inference scaling I recommend Is AI progress slowing down?
16 October 2024
From one piece Graham Hancock: Lost Civilization of the Ice Age & Ancient Human History | Lex Fridman Podcast #449 2 beliefs, in the piece's order there
-
Their words
But what’s not really been addressed before is why that happened, why the Gulf Stream was cut, why a sudden pulse of meltwater went into the world ocean, and it was so much of it and it was so cold that it actually stopped the Gulf Stream in its tracks. And that’s where the Younger Dryas impact hypothesis offers a very elegant and very satisfactory solution to the problem.
-
Their words
So the suggestion is that it wasn’t one impact, it wasn’t two impacts, it wasn’t three impacts, it was hundreds of air bursts all around the planet.
6 August 2024
13 July 2024
19 June 2024
-
Their words
I think if we can achieve that amount of inference compute, where it leads to a dramatically better answer as you apply more inference compute, I think that will be the beginning of real reasoning breakthroughs.
16 August 2023
10 January 2023
13 September 2022
Nothing matches.
What is a korrent?
A korrent is a belief a person has stated in their own words: one sentence stating the claim, backed by a quote and a source, kept at korrents.com.
Under a name here, the quoted block is what they actually said. The korrent beneath it is the claim those words support, in korrents' wording — tap it to see the record, its source, and who else holds it.
Nobody here wrote their own korrents. They are compiled from public statements, and a person can change their mind, which is recorded too.
About the English under a post
Some people here publish in a language other than English. Where they do, this site shows a machine translation beneath the post, in this typeface — the site's own, not theirs.
The post itself is never changed, moved or hidden: what is set in the serif above is exactly what the person published, and it is what to quote them on. A translation can be wrong in ways that matter, especially about tone.
Only the post's own words are translated. A quoted post, a linked article and a belief on korrents.com are left in their original language.