I had seen the dudes wearing bags of nuts and throwing them on the ground, while on my runs. I had no idea what was behind it.

this shortfall accumulates quickly for larger scaleups, with the swarm falling further and further behind. To get the same performance gain as a 100x scale-up of the number of tokens for a single agent you need to scale up the swarm size by 900x to 15,000x.

So why would you ever use swarms?

The most important answer is speed. The 4-agent swarm needed about twice the total number of tokens to get the same performance, but in terms of tokens per agent, it only needed half as many. Since the agents are run in parallel, this means it can theoretically achieve the same task in half the time. And the same was true when moving from 4 agents to 16. In total, one could achieve the task in about 1/4 the time for 4x the cost. In reality, the speedups aren’t quite this good (perhaps because some agents use more tokens than the average), but they are substantial. So for situations in which you’d pay a large premium for speed, swarms can be very useful.

FROM:

Toby_Ord

Swarm Scaling

Source

Communication costs mean super large groups break down in benefit. This is why you need subdivision to get major things done. But scaling up the group means greater expense for faster results.

I used to tell people to look around the room you’re sitting in. Look at everything. Every noun. It almost all exists because a person willed it into existence. Nothing is stopping you from doing the same.

I see people online all the time dismissing someone else’s success because “I had that idea first” or whatever. I mean… yeah? If so then the difference is… you. So a bit of a self own whenever I hear that.

FROM:

Mitchell Hashimoto

Mitchell Hashimoto on X: 'I Can’t Stress Enough How Little an Idea Matters Compared to the Agency of the People Executing the Idea. I Have Had the Privilege of Knowing and Sometimes Even Working With Some of the Most Successful People (By Various Metrics). The Difference Between Mediocre and Excellent' / X

Source

Action is the main difference between people.

AI likes to say vacuous punchy statements like “the final takeaway” or “here’s the rub”. Then contrastive parallelism “connections, not CPU”. Contrastive parallelism should scarcely exist in technical writing, regardless of whether it’s AI generated.

FROM:

energy123

Hacker News

Source

This is often what irks me the most about AI slop writing: the contrastive parallelism is often done through an incongruous comparison - two things that are actually orthogonal or unrelated, not parallel. The LLM constructs the sentence but upon reading it contorts my brain unpleasantly.

rewrite in ASD-STE100 Simplified Technical English

I’ve set up a snippet/shortcut at work so when I type ;ste it gets expanded to ASD-STE100 Simplified Technical English.

Prediction markets are “truth machines,” in the sense that they give people incentives to figure out what is true and bet on it. But they are also, obviously and increasingly, falsehood machines, in the sense that they give people incentives to make bets and then distort reality to make those bets pay off. Because of prediction markets, it is now lucrative to figure out what song will be the most popular, but it is even more lucrative to make a different song look most popular. Prediction markets can make it harder to know what is true.

I guess this is kinda how the stock market and any market works, too? If people can game the incentives, some will, of course, cheat. No market is a truth machine.

Armin theorizes that this is because more recent Anthropic models have been specifically trained (presumably via Reinforcement Learning) to better use the edit tools that are baked into Claude Code. This has the unfortunate effect that other coding harnesses, such as Pi, may find that their own custom edit tools are more likely to be used incorrectly.

FROM:

Simon Willison's Weblog

Better Models: Worse Tools

Source

Oh man, does this mean we’re in a golden age of generalist/open-source model harnesses that will eventuallly deteriorate as the models are more specifically trained for their own proprietary environments?

Or is the discrepancy and difference in harnesses just a temporary path finding our way to the best harness patterns, at which point they converge and become interchangeable like the models beneath them?

Use AI to make things clearer, not longer. Let it sharpen your thinking, not replace it.

FROM:

noslopgrenade.com

No Slop Grenade.

Source

This is a corollary to not letting others do your writing and thinking for you.

I’ve long wondered why some big corp would sponsor him for doing his stuff in the open, rather than doing it within their walls. For things like this Anthropic Fable release, every corp is clamoring to test it but there are nasty strings attached (mandatory data retention policies, etc.). Simon is able to test it for them indirectly and post about it publicly. So he becomes their barometer for things they are too risk-averse to do.

Interesting. I wonder if there’s more to do in that way, or to be more direct about it. Or maybe he is and I’m just not seeing it?

Most often the issue isn’t “empowerment”. People start out wanting to achieve things, to change things to be better, to do a good job. The active issue is disempowerment

Popular security culture often dwells on clever hacks and clean fixes. But for the difficult real-world problems, often there is no right answer or simple fix, only the hard work of building defenses thicker and thicker.

FROM:

The Cloudflare Blog

Mitigating Spectre and Other Security Threats: The Cloudflare Workers Security Model

Source

I’m exploring sandboxing at work ad this was a great read.

People whose brains have been eaten by LLMs still maintain that ‘It’s not gradient, it’s texture’ or whatever is still their idea, expressed by the machine, but there is almost never any idea there at all. If your ideas were any good, you wouldn’t need to use the machine; as it stands your sub-literate scrawlings are the best thing about you. At least they’re yours.

But you people don’t listen. However bad a writer you think you are, you are not worse than AI. But you still keep letting it do your writing for you, as if I won’t be able to tell. Listen: I can tell. I can always tell. You think I won’t notice, but I will. There’s no hiding from me. If you let AI do your writing I will find out, and I will kill you.

FROM:

Sam Kriss

If You Let AI Do Your Writing, I Will Come to Your House and Kill You

Source

I love Sam’s writing.

I want people to enjoy my writing. That’s why I write it myself.

I have taken to having LLMs check my writing and critique it, but I write it myself. At least then, if someone is moved, I moved them.

I will never shut up about the value of living and breathing openness. I learned / gained so much from opening myself up to collaboration AT ALL TIMES. my life is a lot richer now.

FROM:

Todepond dot com

My Year of Liveness

Source

Life is richer with others.

In crisis communication, a well-known formula is: outrage x hazard. So even if the hazard is low, if concern is great, you’d better be speaking with clarity, acknowledging uncertainty, listening to the questions, concerns, and confusion, and bringing people along for the ride.

Calibrate to the audience perception.

There’s a section on ‘patient diaries’, which are not written by a patient but written about a patient. Apparently when you are in a medically induced coma, or treacheotemized, etc, people come back and often have traumatic memories - not from the actual events themselves but from their brain trying to make sense of what happened; they think they were abducted by aliens, or their loved ones were replaced by robots, or they were molested in their incapacitated state. So some Danish nurses started a practice of keeping a handwritten diary for the patient, which their loved ones can also add to, about what’s happening - and then when the patient is back to normal, they give them this handwritten account, and it really helps ground the patient back into reality and make sense of their lost time. It’s so weird, and so beautiful an idea!

FROM:

Ex Libris Kirkland: Recently Added Notes

Note On: The Notebook: A History of Thinking on Paper

Source

Taking notes as medicine.

I didn’t know about this, but already I started a journal for our baby daughter like this. We leave a note for each day and have family/friends leave notes when they are with her those days. The hope was for it to be a keepsake for her, but maybe it will help ground her as a person, to have this history of her infancy.

So while one could in theory be a valid contributor that makes use of LLMs, from the perspective of contributor poker it’s simply irrational for us to bet on LLM users while there’s a huge pool of other contributors that don’t present this risk factor.

The people who remarked on how it’s impossible to know if a contribution comes from an LLM or not have completely missed the point of this policy and are clearly unaware of contributor poker.

For us the ability to provide contributors with an engaging ecosystem where they can improve their systems thinking and interact with other competent, trusted and prolific engineers is a critical aspect of our business model.

As I’ve mentioned before, Zig is able to punch well above its funding class because we put huge effort in thinking about technical, social and business management issues that surround the project.

FROM:

kristoff.it

Contributor Poker and Zig's AI Ban

Source

Contributor Poker is playing the player, not the code/cards.

They’re building a self-sustaining ecosystem of maintainers. That’s what they need you for, not for your one specific contribution. They want your mind, not your code.

From taxes to the environment to public broadcasting like PBS and NPR, the Senate has recently passed record levels of legislation and confirmed record numbers of nominations with senators representing less than half the people.

Using historical data, GovTrack found 56 examples of Senate votes on legislation that passed with senators representing a “population minority.” 26 of those 56 examples, nearly half, have occurred since President Donald Trump’s current term began.

FROM:

Jesse Rifkin

Record Number of Bills and Nominations Passed With Senators Representing a Population Minority

Source

The Will of the fewer and fewer people

We may not be responsible for other people’s choices, but where does what you’re entitled to end and what I’m entitled to begin?

On the same flight, a woman was traveling alone with three small children. I couldn’t help but notice them, because they were all wearing matching outfits, including the woman. Weird. Somewhere over Arizona, she hurried her two older kids, both still in diapers, to the bathroom to change them. When she passed me, she looked at me imploringly. “Can you hold my baby?” She asked, before thrusting him into my arms to chase after the rest of her brood. The baby was chill as hell, with big blue eyes that looked shocked at being so casually pawned off on a stranger. I stood in the aisle rocking him back and forth on my hip, missing the time when my own kids were small enough to rock so easily, feeling flattered someone saw me and assumed I understood the most essential truth: we’re all turbulating through this beautiful, miserable shitstorm together.

FROM:

Hallie Haglund

Flying With Kids and the Future of Mankind

Source

This is how you survive a storm: by helping others and receiving help yourself in mutual aid. Surviving hard things is relational, not individual.

I’m referring to that experience as the Ghost Effect (movie, not CMS). If you talk about something dangerous that might happen in the future, people nod along and file mental notes. If you talk about something dangerous that happened in the past, they listen to see what they can learn. But if you talk about something huge and dangerous happening right now, people get really polite and change the topic.

The combination of vast scope and immediacy—and the fact that you can’t really see or touch what’s happening—means that if you choose to ruin the party by discussing it, you transform into a Patrick Swayze-style ghost. You’re yelling, You’re in danger! You think they can hear and see you, but instead they just feel kind of cold.

FROM:

Paul Ford

The Ghost Effect

Source

People ~hear what they want to hear~ ignore what they want to ignore. They would prefer to ignore the uncomfortable dangers.