Dr Addi Haran Diman

Startup CEO (@a16z @speedrun 006) building multiplayer AI, done right / Former president @OxfordStudents, researched and taught social data science @UniofOxford

[Sahil @sahilypatel]

legendary hacker news comments that aged like milk

[Modaic @modaicdev]

We're announcing Mo-2.0 a frontier System One model with 8 output types for deciding and extracting.
- choice
- noul
- score
- box
- span
- text
- number
- datetime

[Aaron Levie @levie]

Agent swarms have the potential to drive yet another leap in output from AI. The impact will be a clear separation between teams that adopt this pattern of work and those that don’t.

The progression has roughly been: chat gave us quick turn AI responses that let us work a bit faster, but we were still rate limited by our interactions; then agents that can go off and complete entire tasks for us, accelerating hours or days of work into minutes; to now swarms of agents that are goal seeking and will find any route to the successful completion of a project, where there’s no inherent limit to the productivity gains.

We can imagine a future where, for any task in which value can be created within a self-contained digital environment, swarms will be able to complete the work faster than any other approach.

Essentially any area where value creation is currently bound by the amount of research you can do, the amount of decisions made with a set of software tools, the amount of data processed or consumed, and the amount of code (or code adjacent) that can be generated.

This means swarms will play a big role in game development, coding, trading, life sciences research, IT automation, cyber, math breakthroughs, animation, and plenty of other areas of work that have these characteristics.

There’s a lot of opportunity to build the harnesses, applied AI products, and infrastructure for how these agents operate. And most importantly there’s a ton of opportunity for those that can wire up work in this way. It’s going to be an entire new field of FDE work available.

[Dr Addi Haran Diman @AddiHD02]

Swarm riding will be the new knowledge work. It’s a science and an art, and it’s always a marvel to watch (ie just looking at how @emax works)

[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?

[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?

[Gagan @GaganSKhurana]

Insurance is the single most important market on planet Earth, yet it still runs on manual work. Today ancep is coming out of stealth with an $8.5m pre-seed round to change that. Our AI agents are being used daily by the largest global brokers to help them place more risk.

[Justin Lo @loresjpg]

Nice DeepSWE score! Does it survive a lowercase filename?

Introducing MaxxBench: 89 DeepSWE tasks with small changes to the instructions or starting code. Same engineering problem, slightly different requirement.

Models clustered around 74% on DeepSWE end up 23 pp apart.

Good at coding, or good at DeepSWE? 🧵

[james hawkins @james406]

i can't say much, but these coming weeks will change everything

you're really not ready for what's coming

well done to the lab that made this happen

There. I made a sufficiently vague post about something big happening. Want me to make it even more vague, or lean in to the fear angle? Also, while I have you, did you still want to order that Hello Kitty-themed mechanical keyboard?

[George Jefferson @GeorgeJeffersn]

Reselling shoes at 15 years old changed my life

The UK gov should give young people more opportunities to make money, not less

[Olivia Moore @omooretweets]

Sometimes you see a graph that stops you in your tracks

October makes September look like a rounding error 👀

Never seen anything like this in my decade as an investor…

[Omar Ghandour @omarships]

a million views for an mcp server with tooling that writes notes.

that’s the feature. your agents write down the slop they think is important, and we’re calling it a shared brain.

the part people missed is that it’s metered lol. shipped in the same release as the credit system. 1,500 standard credits, 500 select, admins buy more when you run out.

but the price isn’t the problem. if it was an actual brain it would be worth the money. this is a gimmick.

people have done this with obsidian for over a year. another 1,100 open source repos on github for shared md files.

notes are authored. a brain is derived.

a brain is your company rebuilt as an ontology: customers, invoices, threads, renewals, people, and how they connect. the charge in stripe, the thread in gmail and the renewal in the crm resolve to one customer. every fact carries where it came from and when it stopped being true. it’s what palantir hands a government, pointed at a company your size.

a notes file has none of that. no source, no expiry. agent A guesses wrong on tuesday, agent B reads it as fact on wednesday, by friday four agents agree on something nobody checked.

and what @NotionHQ shipped leans on every spun up chat to decide what goes into your “brain”. the agent picks, the tooling stores. so the brain is only ever as good as whatever that session happened to notice.

is it useful if you have nothing of the sort? barely. chat memory has gotten very good on its own. and the people running cross-agent work, which the video is clearly aimed at, are well past the point where this helps them.

what gets me is the people actually building brains. cheap, effective, scalable, capturing ontology no matter which tool it came from, now lumped in with a notes file because of echo chambers and bots pushing it. now they spend their week explaining why they aren’t a folder.

that gap is where most agent failures of the next two years come from.

[Dr Addi Haran Diman @AddiHD02]

Biggest green flag to look for when interviewing engineers: they became hackers EARLY

[Jordan @justmazer]

average age founders started coding vs. school

(data from our @speedrunalpha pipeline answers)

[Dr Addi Haran Diman @AddiHD02]

All agent architecture should be event-driven, let's kill cron jobs

[AgentMail @agentmail]

Your agent can now get its email events without a server.

Polling webhooks are live on AgentMail, built with @SvixHQ.

The agent fetches new events whenever it runs.

Built for agents in sandboxes, serverless functions, cron jobs, or on a laptop.

[Dr Addi Haran Diman @AddiHD02]

The short term impact of this capability is democratising tinkering with small models - produce a small neural network with the same ease as a ghibli picture.

Is this what led to the worry about “emergent RSI capability in new models” from the labs back in the summer?

[Alex Krentsel @AlexKrentsel]

i've reproduced the results from the MIT blog post today on weight-space programming. but one additional result i have to add: this is a new, step-change emerging capability that's appeared in just the latest set of models.

[Marcus Segal @marcussegal]

There has never been a better time to build and @speedrun is here to help. Apply today at https://t.co/4XOVvSQyi1

[Josh Lu @JoshLu]

Don't Build Alone

As a @speedrun partner, some of my favorite conversations with founders starts with: “here's a thing i'm stuck on...”

Maybe that thing is pricing. Maybe a promising customer has gone quiet. Maybe the founders are hiring their first salesperson and trying to figure out what good looks like.

Working these things through with founders is one of my favorite things about this job.

At Speedrun, we want to be your earliest backers and work alongside you through those decisions. You get our time, our attention, and a team of world class operators who can help you work through what’s blocking you.

Building something great is hard - don't feel like you have to do it alone. Let us work with and support you on the journey ahead.

SR008 applications close this Sunday at midnight.

If you’re building, we'd love to hear from you. Application link below

[Ryan Peters @ryanirl]

A small swarm versus a single agent in an autoresearch harness

[Seb Johnson @SebJohnsonUK]

The UK budget is in. Here's everything you need to know:

🇬🇧 No exit tax or increase in capital gains

🇬🇧 Brand new "founder’s relief” - 10% tax up to £10m lifetime gains and BADR thresholds cut by half

🇬🇧 The non-compete ban we already knew about, but slightly more fleshed out

🇬🇧 The additional AI money, which we already knew about, but in more detail

🇬🇧 Some very important changes to visas to attract better talent to the UK

🇬🇧 £5bn for BBB including more money to back female founders

🇬🇧 "Young Entrepreneurs Allowance" which covers living costs for young entrepreneur

I have probably missed some bits but overall this is a lot more pro-tech and pro-entrepreneur than I think people were expecting.

Great stuff from @adepledge, @Dom_Hallas and all those working tirelessly behind the scenes to advocate for a strong entrepreneurial economy here in the UK.

LETS GO

[Jacqueline Young @jyoung008]

Every company needs capital to scale and grow. To-date, @speedrun companies have raised nearly $1B.

If you want a deeply experienced team in your corner every single step of the way as you navigate early-stage venture, fundraising, and financing, drop an application!

Can't wait to work together 🚀

[nader dabit @dabit3]

Biggest opportunity now as a software engineer is clear: master parallelism → outsize scale of output. Value creation in software has always come from leverage: small input, outsized output.

Parallelism-maxxing is not extremely quantifiable because it's so new. The spectrum runs between people managing a couple of agents on their local machine to those running an infinite number of agents + computers at a time than can be spun up and down on demand. You want to be at the far end of this spectrum, and you should be ROI-maxxing every token spent.

If you can do this and prove it, you will have an infinite number of opportunities available.

If you're still trying to do this locally you are at an objective disadvantage, equivalent of hosting a startup on a server in your closet vs using AWS. And if you're spending your time building and maintaining infrastructure that gives you no competitive edge, the AWS analogy applies in the question "why are you wasting your time doing undifferentiated heavy lifting instead of building unique customer-facing features?".