Rhys

Exploring what’s next. Helping AI communities as I find my next role. Meliorist. Prev: | @roote_ | @mitdci | @EthereumDenver

[Rhys @RhysLindmark]

11/ Amazing new benchmark highlighted by @simonw

Felony Bench 😂

https://t.co/rjiJ2CDMVD

[Rhys @RhysLindmark]

1/ Thread on @aiDotEngineer

First, @cjav_dev from Anthropic on Machine Trust.

Two main takeaways:
- Try to mitigate the blast radius of agents gone wrong.
- Trust is so contextual. Some people want to allow computer use on Fidelity to buy & sell stocks, some don't.

[Rhys @RhysLindmark]

10/ This @Steve_Yegge talk is so fucking good. So funny.

Steve has been making these amazing comics on AI-human office dynamics. Like Dilbert for 2026.

Link here: https://t.co/B84cWOQLQB

[Rhys @RhysLindmark]

10/ This @Steve_Yegge talk is so fucking good. So funny.

Steve has been making these amazing comics on AI-human office dynamics. Like Dilbert for 2026.

Link here: https://t.co/B84cWOQLQB

[Rhys @RhysLindmark]

The @nullfellows team is cooking. Accelerating human capital in Europe.

Love their reflection on the journey:

"You don’t solve talent with a funnel.

You do it by hand, by actually getting to know people and being the first ones to take them seriously. Being someone’s first champion is the biggest opportunity there is.

This is the most exciting time in history, and it belongs to young, capable and ambitious people."

Worth applying, stopping by the office in Helsinki, or following them for more. @Daniel_Almi @ArtemKislukhin @proactivepath

{}

[Rhys @RhysLindmark]

@exgenesis @comm_archive Others I want to see:

- agentic
- vibe coding (I think you may have done this in a slightly diff form already)
- slop
- effective altruism
- e/acc
- "it's time to build"

[Rhys @RhysLindmark]

@exgenesis @comm_archive holy shit this is so cool

[Rhys @RhysLindmark]

@hugbubby @tamaybes Interesting to check in on these!

1. MMLU >80%: Yes, Sep 2024
2. $1B of compute: Yes, Feb 2025
3. MATH >90%: Yes, Sep 2024
4. APPS >80%: Yes, Feb 2025
5. IMO gold: Yes, Jul 2025
6. Dishwashing robot: No (not close)
7. Tesla FSD: No (close)

[Alexander Berger @albrgr]

This newsletter is my new favorite source for easily following AI developments: https://t.co/GthSfXaCYE props to @peligrietzer et al

[Rhys @RhysLindmark]

@albrgr @peligrietzer Yes! Makes me think of this other excellent newsletter

The @NunoSempere extended epistemic universe :)

https://t.co/XceSQwVoCr

[Erik Torenberg @eriktorenberg]

Some people need to get their foot in the door at the right place and rise up by doing great work rather than be overly precious re: scope and title at the outset. What’s most important is where you finish, not where you start. Prove yourself.

[❤️‍🔥 xiq @exgenesis]

"you can just do things"

how did a tpot meme end up mentioned by the president of El Salvador and the Department of War?

first @comm_archive quick data report on how ideas spread

[Rhys @RhysLindmark]

@maxsbennett @AsteraInstitute @dileeplearning Yusssss

Great fit!

[Eryney @eryney_ok]

Unironically Shelby is the single best, most insightful VC in biotech. There are several ideas that I read in her /compound blog, that I then saw unfold 6 months - 1 year later.

[David Lang @davidtlang]

BIG NEWS!

For the past few years, we've given scientists budgets so they can fund their peers with small, fast grants. Today, we're going bigger.

We're opening up a call to find and fund emerging "field builders"—people with the ambition and curiosity to kickstart a new field of science and technology. You'll get a budget of $70,000 to regrant (up to $10,000) out to projects and ideas and join a community of explorers.

Do you have an idea for a new field of science that *should* exist? Please apply!

[Matt Huang @matthuang]

Swarm riders and Swarm-native companies

AI-use even among highly AI-native people is power-law distributed.

There seems to be a new class of people — call them “swarm riders” — who are capable of consistently commanding large swarms of agents to productive ends.

This is a new kind of managerial skill, analogous to the skill it takes to manage large organizations of people, but wholly different in nature and of course more technical.

Based on the swarm riders we know (including many at @Tempo and @Paradigm), these people are capable of productive token spend that is 1000x-10000x higher than median software engineers, and 100x higher than top 5% Claude Code users.

The art of swarm riding is different from software engineering and more akin to machine learning, where practitioners grapple with black box systems: defining reward functions and evals, transmuting non-verifiable problems into verifiable problems, running experiments, etc.

Swarm-native companies are a new class of company founded by and populated by swarm riders. These companies are organized around swarms in a way that conventional companies will struggle to replicate, as the cost of internal human coordination is sand in the gears of an effective swarm. Effective swarm-native companies will define the right reward functions and evals, remove functional and bureaucratic barriers, and then let the swarms cook.

We’ve gotten to see a few swarm-native companies up close — reshaping financial rails @Tempo, building a new gaming platform @Spawn, cyber swarms at Cathedral Systems — and it’s clear these teams wield a significant advantage over pre-AI teams.

Some questions I’ve been wondering about:

1) How fast will swarm riding diffuse through the economy? The implications for token spend could be significant.

2) Can swarms be effectively adopted by existing companies or will it generally track the formation of new swarm-native companies?

3) What becomes scarce as swarm capabilities themselves commoditize (through frontier models becoming better at swarming, swarm capabilities built into harnesses, etc)? Cultures and teams that can define reward functions well and otherwise get out of the way?

[kasey @kaseyklimes]

imagine if every decision you made had a guardian agent that defends it on your behalf across your organization — in every agent session, slack thread, and meeting

if, instead of repeating yourself constantly (to people and to agents), your decisions were enforced through infrastructure, along with the rationale behind them so learning compounds

if you could actually trust your agents to ship on your behalf, with your entire team's judgement as their guardrails

if you could trust new engineers and designers to build in a codebase they haven't learned yet, because the decisions already made in it serve as a safety net around their agents (and teaches them that tribal knowledge along the way)

and imagine this realtime merge conflict for decisions happens with zero change to anyone's workflows. no new app, nothing to maintain, installed with a single prompt

this is what we've built at @primitive_hq and its going to change organizations forever

[Rhys @RhysLindmark]

@KeithZhai @aiDotEngineer @cjav_dev He didn't!

It was just a little example of differing expectations for trust.