Latent Space: AI News & Views(@latentspacepod)

@davis7 @benhylak @dexhorthy @allisonology @altryne @swyx @OpenAL find the High Taste Testers wherev...

Latent.Space (@latentspacepod) — Fri, 16 Oct 2026 23:54:23 GMT

@davis7@benhylak@dexhorthy@allisonology@altryne@swyx@OpenAL find the High Taste Testers wherever the LS pod network airs: youtube.com/watch?v=ZOH8Q4…

💬1🔄0❤️2👀1801📊2

🆕 show: HIGH TASTE TESTERS We convene top Al power users to compare what actually works-and what ...

Latent.Space (@latentspacepod) — Fri, 16 Oct 2026 23:53:50 GMT

🆕 show: HIGH TASTE TESTERS

We convene top Al power users to  compare what actually works-and what falls apart once you start using it.@davis7, @benhylak, @dexhorthy, @allisonology, @altryne and @swyx give you unfiltered takes on @OpenAl DevDay, Dots vs Muse vs GrokBot vs Instinct, who's winning coding, personal assistants, and the gap between impressive Al capabilities and useful products.

From Pi-powered personal agents and decision models to subagents, context management, prompt injection, and assistants that can spend money: what should you actually use?

CHAPTERS
00:00 Introducing High Taste Testers
01:26 OpenAl DevDay: Dots, Spaces, and product execution
10:28 Great models versus great products
14:39 Agents versus assistants
18:08 Building a personal assistant with Pi
21:07 Decision models, Jev, and predictive interfaces
32:46 Model speed, writing quality, and subagents
37:29 Agent loops, forks, and context management
43:19 Haiku, Muse Spark, and Grok
50:00 Local models and coding benchmarks
54:09 Prompt injection and assistants taking action
57:50 Tools, payments, and assistant overload
1:02:16 Delegation, computer use, and everyday workflows
1:10:48 Al hardware and closing thoughts

THANKS TO THE TESTERS
Ben Davis - @davis7
Ben Hylak - @benhylak
Dex Horthy - @dexhorthy
Allison Durham - @allisonology 
Alex Volkov - @altryne

Your browser does not support the video tag.

💬15🔄13❤️97👀32326📊47

A big theme at the @aiDotEngineer New York conference this week was how to balance agent autonomy wi...

Latent.Space (@latentspacepod) — Wed, 14 Oct 2026 21:58:51 GMT

A big theme at the @aiDotEngineer New York conference this week was how to balance agent autonomy with security and safety. @OpenAI, @AnthropicAI, @twosigma & @Bloomberg focused on this issue in their presentations. The two frontier labs were keen to talk up the threshold their latest models have crossed, but safety was top of mind for all. latent.space/p/crossing-the…#AIEngineerNY

💬4🔄0❤️9👀1288📊6

From Wet Lab to RL Env: 10,000 Drug-Discovery Campaigns in a Weekend @LeashBio’s @allmeasures expla...

Latent.Space (@latentspacepod) — Wed, 14 Oct 2026 20:44:48 GMT

From Wet Lab to RL Env: 10,000 Drug-Discovery Campaigns in a Weekend

@LeashBio’s @allmeasures explains how DNA-encoded libraries generate billions of protein–molecule measurements, why no team in Leash’s Kaggle competition beat chance on an unseen chemical library, how public datasets can teach AI to recognize the chemist instead of understand the chemistry, why engineering protein mutations exposes model memorization, and how replayable experimental worlds let agents run over 10,000 medicinal chemistry campaigns in a weekend. Plus: Hermes vs. Boltz, moving from DEL hits to mouse studies in six months, and why experimental rigor and the Bitter Lesson of scaling could unlock a new generation of AI-driven drug discovery.

Your browser does not support the video tag.

💬5🔄5❤️24👀3113📊12

Full Youtube Episode: https://t.co/OO6VQe02Xq

Latent.Space (@latentspacepod) — Tue, 13 Oct 2026 17:42:42 GMT

Full Youtube Episode: youtube.com/watch?v=BHiYom…

💬0🔄0❤️1👀825

Robots That Predict the Future: Dyna-2, Million-Hour World Models, & the Bitter Lesson of Robotics ...

Latent.Space (@latentspacepod) — Tue, 13 Oct 2026 17:42:41 GMT

Robots That Predict the Future: Dyna-2, Million-Hour World Models, & the Bitter Lesson of Robotics

@DynaRobotics’s @JasonMa2020 explains how world-action models predict the future to guide physical robot actions, why training on over a million hours of human video lets robots adapt to new tasks with as little as 45 minutes of demonstrations, how synthetic data and self-improving world models could accelerate robotics, why agentic harnesses are enabling hour-long autonomous workflows, and how the Bitter Lesson of scaling, full-stack hardware, and real-world deployments at Din Tai Fung are turning physical AI from impressive demos into useful robots.

Your browser does not support the video tag.

💬8🔄12❤️36👀5029📊17

At AI Engineer New York, Anthropic's @cjav_dev & @HarrisonStall from @ClaudeDevs drew on lessons...

Latent.Space (@latentspacepod) — Mon, 12 Oct 2026 18:32:57 GMT

At AI Engineer New York, Anthropic's @cjav_dev & @HarrisonStall from @ClaudeDevs drew on lessons from Unix and Linux to explain how agents can tackle bigger projects without overwhelming developers. https://t.co/4i30WOQdJs

💬7🔄3❤️23👀6853📊14

New Show Announcement: High Taste Testers Good food and good ai convos Pilot episode 🔜 https://t....

Latent.Space (@latentspacepod) — Mon, 12 Oct 2026 15:29:58 GMT

New Show Announcement:

High Taste Testers
Good food and good ai convos

Pilot episode 🔜youtube.com/@latentspacepod4 

ft: @davis7, @dexhorthy@AdjectiveAllison, @altryne, @benhylak on DevDay/Dots, Jev, Muse, Grok, Claude v GPT v Gemini, and all the hot topics du jour!

Your browser does not support the video tag.

💬2🔄3❤️22👀7810📊7

Why AlphaFold Didn't Solve Protein Folding: The Bitter Lesson of Biology, Virtual Cells, & Curing Al...

Latent.Space (@latentspacepod) — Sat, 10 Oct 2026 00:32:51 GMT

Why AlphaFold Didn't Solve Protein Folding: The Bitter Lesson of Biology, Virtual Cells, & Curing All Disease latent.space/p/biohub-deepm…@GoogleDeepMind’s @pushmeet and @biohub’s @salcandido explain why AlphaFold's breakthrough hasn't solved protein dynamics, why scaling AI compute and data alone isn't enough to understand biology, how protein language models contain hidden knowledge about structure, function, and evolution, why future frontier models may understand other AI systems better than humans, how better biological data and virtual cells could unlock the next wave of breakthroughs, and what it will take to achieve 10–100x faster drug discovery on the path to curing all disease.

Your browser does not support the video tag.

💬7🔄7❤️31👀7819📊16

It's true we no longer need to keep our laptops open, but for all the recent progress of OpenAI and ...

Latent.Space (@latentspacepod) — Wed, 07 Oct 2026 14:33:46 GMT

It's true we no longer need to keep our laptops open, but for all the recent progress of OpenAI and Anthropic, cloud-based agent harnesses aren't fully solved yet.

Maybe we need the creators of Kubernetes to tackle the problem? Stacklok is built by two of them, @cmcluck & @jbeda, and they aim to bring agent harnesses fully to the cloud.

We talk to McLuckie and Beda about their "cloud-native harness": latent.space/p/stacklok

💬7🔄0❤️13👀1424📊9

More from @trq212 on local hands and how it is being implemented in Claude Code: (see more on htt...

Latent.Space (@latentspacepod) — Tue, 06 Oct 2026 05:20:34 GMT

More from @trq212 on local hands and how it is being implemented in Claude Code: 

 (see more on latent.space/p/thariq)

Your browser does not support the video tag.

💬3🔄4❤️90👀85352📊30

Inside-Out AI: change how the company works, then what customers experience. After leading Meta’s L...

Latent.Space (@latentspacepod) — Fri, 02 Oct 2026 15:06:57 GMT

Inside-Out AI: change how the company works, then what customers experience.

After leading Meta’s Llama models, @Airbnb CTO @Ahmad_Al_Dahle is putting that approach into practice — incl w/ Everest, an internal tool that helped speed up a product launch. latent.space/p/airbnb

💬3🔄3❤️11👀1756📊6

Recursive Language Models: Claude Code, Agent Swarms, Big Research Bets, & the $40M AI Problem Solve...

Latent.Space (@latentspacepod) — Fri, 02 Oct 2026 00:32:32 GMT

Recursive Language Models: Claude Code, Agent Swarms, Big Research Bets, & the $40M AI Problem Solver latent.space/p/rlm

MIT’s @a1zhang explains why Claude Code, Codex, and Pi are basically the same, how RLMs use code, context offloading, and recursive subagents to generalize across tasks, why one expert can sometimes replace a trillion-token brute-force search, what OpenAI’s 10,000-agent, 130B-output-token experiment reveals about the future of language models, and why academia’s biggest advantage is the freedom to take weird, ambitious research bets.

Your browser does not support the video tag.

💬17🔄10❤️74👀9278📊41

.@AnthropicAI’s Thariq on what Claude Mods unlock: conversation context, subagents, structured outpu...

Latent.Space (@latentspacepod) — Thu, 01 Oct 2026 19:47:01 GMT

.@AnthropicAI’s Thariq on what Claude Mods unlock: conversation context, subagents, structured output, and control over the UI.

“It has a bunch of things in the scope … like how many turns is in this conversation, right? Like how many tokens have been used?”

“Because it’s all happening in process, you can spawn subagents … you can parse the results of those, you can use structured output to sort of like return them.”

“And then you can modify the UI, which you can never do in hooks.”

@trq212@ClaudeDevs

Your browser does not support the video tag.

ClaudeDevs@ClaudeDevs

You can now mod Claude Code:

- Change how it behaves
- Customize the UI
- Swap in your own features

Write one with a few lines of TypeScript, or have Claude build it for you. Mods ship inside plugins, so you install them with /plugin in the CLI or desktop app.

A few examples:

Your browser does not support the video tag.

💬7🔄7❤️30👀7114📊11

Full Episode: https://t.co/aXe3Ta19E8

Latent.Space (@latentspacepod) — Thu, 01 Oct 2026 19:47:01 GMT

Full Episode: youtube.com/watch?v=IZAlq-…

💬0🔄0❤️2👀861📊1

more!

Latent.Space (@latentspacepod) — Wed, 30 Sep 2026 23:50:20 GMT

more!

Sabrina Halper@SabrinaHalper

NEW episode with @alexatallah, co-founder of @OpenRouter. He recently signed an agreement to be acquired by Stripe for a reported ~$7B+.

This is his first interview since the news.

This one is a particularly full-circle moment for me. Alex was the first person I interviewed, back in early 2021, when I had started a podcast my senior year of college and he was building OpenSea.

Five years later, Alex has built another incredible company. 

We talked about why OpenRouter and Stripe make sense together, and their shared belief that the future shouldn’t be dominated by one model or one giant lab. Alex thinks there could eventually be thousands or tens of thousands of models, and OpenRouter is building the marketplace and infrastructure to help people and agents choose between them. 

What we got into:
• Why OpenRouter saw a real acceleration in AI usage around January 1, 2026, and why Alex told his company it was time for “no more autocomplete” and to move to agents. 
• Why he thinks AI agents could be tied to most GDP growth in the near future, as agents start buying things and transacting on our behalf. 
• An anonymous Chinese model becoming the biggest model launch in OpenRouter’s history, and what that says about how much model brands actually matter. 
• Why the cost of an AI model will increasingly be measured by how much it costs to complete a task, not how much each token costs. 
• How the internet itself may start being built for agents rather than humans. 
• What humans will still want to own and create as agents take on more knowledge work, software work, and eventually physical work through robotics. 
• And how Alex finds company ideas by looking for the bottlenecks between what models can do today and what they’ll likely be able to do 12 months from now. 

Alex is a deep thinker and a one-of-a-kind company builder. I feel very lucky that I got to interview him at the beginning of my career, and even luckier to do it again now.

Hope you enjoy this one :)

Your browser does not support the video tag.

💬0🔄1❤️1👀1245📊1

OpenAI's New Agent Stack: Superhuman Computer Use, Dots, Decisions API, UltraFast, & the AI Cloud ht...

Latent.Space (@latentspacepod) — Wed, 30 Sep 2026 23:37:53 GMT

OpenAI's New Agent Stack: Superhuman Computer Use, Dots, Decisions API, UltraFast, & the AI Cloud latent.space/p/devday-2026@OpenAI’s @AriX and @nikunjhanda explain why Computer Use is already becoming faster than humans at real-world tasks, how Dots gives agents their own cloud computers, why better models can now debug and recover from failures, how Decisions API went from idea to product in weeks, and why async tools, UltraFast inference, caching, compaction, and higher-level agent primitives are turning OpenAI’s API into an AI-native cloud.

Your browser does not support the video tag.

💬10🔄14❤️100👀25326📊37

@AnthropicAI @trq212 great notes! https://t.co/JcB1PBmjN3

Latent.Space (@latentspacepod) — Wed, 30 Sep 2026 19:58:00 GMT

@AnthropicAI@trq212 great notes! x.com/yan5xu/status/…

yan5xu@yan5xu

some notes of mine after this ep

1. prompting is two things: does your mental model match, and do you have enough domain vocabulary
not about writing longer instructions. you need an accurate read on what the model can one-shot vs what has to be broken into steps. and you need the words for the problem domain. non-designer says "give me 8 mockups", designer says fonts, ref sites, components. the gap is vocabulary.

2. brutal paradox here
lots of people want to use AI for stuff they don't understand. but if you don't have the domain vocab, the people who actually do the work will out-prompt you. vocab comes from having done it. AI gets you to 60-70. getting from 70 to 95 still needs the domain knowledge.

3. MCP is just a headless API. Artifact is a short-lived UI. they compose across MCPs.
brain / hands / surface reads to me as the agent-native version of classic software architecture. if this works out, building software goes from writing frontends to writing MCP servers, and agents compose the UI. basically: build a piece of software for agents to use, wrap the business logic inside.

💬0🔄1❤️1👀669📊1

we try!

Latent.Space (@latentspacepod) — Wed, 30 Sep 2026 19:57:41 GMT

we try!

Aleix Prat@aleixpratl

lo de la newsletter de @latentspacepod qué es? llevaba mucho tiempo buscando una newsletter que realmente aportara valor en temas de IA, y esta es muy muy buena

💬0🔄0❤️4👀1642📊1

good notes! https://t.co/G9Xeb8YGIf

Latent.Space (@latentspacepod) — Tue, 29 Sep 2026 14:11:33 GMT

good notes! x.com/claude_code/st…

Claude Code Community@claude_code

Worthwhile interview from @trq212 on @latentspacepod 

Practical Tips:

1) We're moving from a single-player game to multiplayer.

Use Artifacts more. 
They're shared memory context for teammates, other threads, and subagents. 

How? 
Just ask claude to make it into an artifact.

2) "Explain using the big picture in few words"  is shockingly more useful than reading through piles of text.

try the plugin `/eli5` it's basically this.

3) Your "unknown unknowns" are the bottleneck. The fix is to try more things and develop intuition & taste:

"To have taste, you have to eat. You have to try a lot of things, iterate, figure out what you want and like, and build a vocabulary for the domain. Then your prompt can bring all of that together." paraphrasing @jxnlco

💬0🔄0❤️1👀867

@AnthropicAI @trq212 https://t.co/aXe3Ta19E8

Latent.Space (@latentspacepod) — Tue, 29 Sep 2026 01:48:58 GMT

@AnthropicAI@trq212youtube.com/watch?v=IZAlq-…

💬0🔄0❤️5👀1683📊1

The Future of Claude Code: Mods, Mutable Software, Pacing the Frontier, & Multiplayer Agents with Th...

Latent.Space (@latentspacepod) — Tue, 29 Sep 2026 01:30:00 GMT

The Future of Claude Code: Mods, Mutable Software, Pacing the Frontier, & Multiplayer Agents with Thariq Shihipar latent.space/p/thariq@AnthropicAI’s @trq212 explains why prompting is still one of the highest-leverage skills in agentic coding, why Claude.md may eventually disappear, how Claude Mods let developers customize the entire Claude Code harness, why mutable software could change how applications are built, how multiplayer agents and Claude Tag are becoming an organizational harness, and why agents hacking Hugging Face exposed a much bigger security problem as frontier models become increasingly capable.

Your browser does not support the video tag.

💬29🔄9❤️76👀37770📊56

STOP making "Jevbench"es, stop asking for public benchmarks, they completely miss the point of Jev a...

Latent.Space (@latentspacepod) — Mon, 28 Sep 2026 22:04:47 GMT

STOP making "Jevbench"es, stop asking for public benchmarks, they completely miss the point of Jev and you won't believe how easy it is to game every benchmark you hold dear

This is @CompleteSkeptic's bitterest lesson of all: picking the right task beats everything

Your browser does not support the video tag.

💬15🔄5❤️100👀87488📊37

Why Stripe Acquired OpenRouter: 10T+ tokens/day, model routing, Stripe, & the $10T token economy htt...

Latent.Space (@latentspacepod) — Fri, 25 Sep 2026 23:30:09 GMT

Why Stripe Acquired OpenRouter: 10T+ tokens/day, model routing, Stripe, & the $10T token economy latent.space/p/openrouter@OpenRouterAI co-founder @alexatallah and @AnjneyMidha explain why betting against a single-model future became one of the most important infrastructure bets in AI, how OpenRouter grew from the Llama and Mistral era into a neutral distribution layer for developers, why model labs can spend billions training checkpoints and still struggle with distribution, how Midjourney and Discord shaped their thinking, why Stripe acquired OpenRouter, and why agentic fraud could become a defining security problem for a multi-trillion-dollar token economy.

Your browser does not support the video tag.

💬16🔄16❤️75👀32085📊41

Full Youtube Episode: https://t.co/NkAySQFkyA

Latent.Space (@latentspacepod) — Fri, 25 Sep 2026 23:30:09 GMT

Full Youtube Episode: youtube.com/watch?v=dCX4PE…

💬1🔄2❤️5👀2472📊3

[AINews] The Future of Latent Space https://t.co/ow9M0KBxbZ - Plans for AINews v3 - Plans for a ne...

Latent.Space (@latentspacepod) — Fri, 25 Sep 2026 05:46:36 GMT

[AINews] The Future of Latent Space

latent.space/p/ainews-the-f…

- Plans for AINews v3
- Plans for a new home!
- We are open for business - and @supabase are our first sponsors!

💬3🔄1❤️15👀27747📊7

Runway’s World Model Bet: Video, Robotics, Real-Time AI, & the Neural OS https://t.co/puu2u5FU9Z @r...

Latent.Space (@latentspacepod) — Fri, 25 Sep 2026 01:40:00 GMT

Runway’s World Model Bet: Video, Robotics, Real-Time AI, & the Neural OS latent.space/runway@runwayml Co-founder & Co-CEO @agermanidis explains why the endgame of AI video is a world model, how Sora triggered an existential sprint inside Runway, why scaling video models may be enough to learn physics, how real-time generation could turn world models into robotics simulators, why the future of software may be interfaces generated directly as pixels instead of HTML and code, and how Runway is building toward a fully neural operating system.

Your browser does not support the video tag.

💬6🔄2❤️12👀1872📊9

This is tracking to be the #1 podcast we've released in all of 2026. We asked Jev's creator @Comple...

Latent.Space (@latentspacepod) — Thu, 24 Sep 2026 15:16:00 GMT

This is tracking to be the #1 podcast we've released in all of 2026.

We asked Jev's creator @CompleteSkeptic to explain Jev, and why he does NOT call it a "Decision Model" — why it's called Jev

available wherever the finest podcasts are sold! reviews help us a lot to keep us relevant in the rankings 🙏

Your browser does not support the video tag.

Latent.Space@latentspacepod

Jev and the System One Model: RLCD, intelligence/$, reliable AI, & the end of chat-first AI latent.space/p/jev@typesafeai CEO @CompleteSkeptic explains why AI can solve extraordinarily hard problems yet still fail to automate basic work, why Jev is built for reliable decisions inside software instead of chat, why TypeSafe rejects public benchmarks and refusals at the API layer, why data and the right task matter more than brute-force compute, how System One Models could reshape coding agents and software, and why even with $1 billion he wouldn’t pre-train a model from scratch.

Your browser does not support the video tag.

💬5🔄10❤️58👀15931📊16

Jev and the System One Model: RLCD, intelligence/$, reliable AI, & the end of chat-first AI https://...

Latent.Space (@latentspacepod) — Mon, 21 Sep 2026 22:14:27 GMT

Jev and the System One Model: RLCD, intelligence/$, reliable AI, & the end of chat-first AI latent.space/p/jev@typesafeai CEO @CompleteSkeptic explains why AI can solve extraordinarily hard problems yet still fail to automate basic work, why Jev is built for reliable decisions inside software instead of chat, why TypeSafe rejects public benchmarks and refusals at the API layer, why data and the right task matter more than brute-force compute, how System One Models could reshape coding agents and software, and why even with $1 billion he wouldn’t pre-train a model from scratch.

Your browser does not support the video tag.

💬62🔄49❤️454👀415331📊220

Full Youtube Video: https://t.co/qrUWmZ5135

Latent.Space (@latentspacepod) — Mon, 21 Sep 2026 22:14:27 GMT

Full Youtube Video: youtube.com/watch?v=cFx9Z3…

💬0🔄5❤️30👀6239📊14

Underwriting Superintelligence: AIUC-1, Agent Audits, Insurance, & the Watchdogs of AGI https://t.co...

Latent.Space (@latentspacepod) — Wed, 16 Sep 2026 20:00:00 GMT

Underwriting Superintelligence: AIUC-1, Agent Audits, Insurance, & the Watchdogs of AGI latent.space/p/aiuc

AI Underwriting Company cofounder @RuneKvist explains why risk and trust may become the real bottlenecks to AI adoption, how AIUC stress-tests agents for jailbreaks, hallucinations, and data leaks, why standards and insurance need to evolve together, what happens when a $20 AI agent creates $200M of liability, and why even after AGI there’s one job the labs can never do themselves: be their own watchdog.

Your browser does not support the video tag.

💬15🔄2❤️20👀4333📊21

Congrats on the raise! https://t.co/yMqg9CapOh

Latent.Space (@latentspacepod) — Wed, 16 Sep 2026 20:00:00 GMT

Congrats on the raise!
x.com/RuneKvist/stat…

Rune Kvist@RuneKvist

The Artificial Intelligence Underwriting Company has raised $55M to audit and insure frontier AI models.

Insurance produces audits the public can trust and understand. The mechanism is simple: insurers need audits to price the risk. They pick auditors they believe will help them price the risk most accurately. Insurers then turn the audits into a price signal that cuts through political debates. Insurers are aligned with the public: if they get the audits wrong, they go out of business. That’s different from today, where labs pick their own auditors, creating public distrust.

Insurance only works if we scale up auditing, and we are building the infrastructure to do that. Millions of agents are being deployed; they leave terabytes of traces and will create thousands of incidents across dozens of risk categories. To scale auditing, we need reliable agents to surface evidence traces for human review.

History guides us. Take cars: auto insurers funded crash tests to price risks; crash tests led us to airbags, lowering the risk. Take electricity: home insurers funded lightbulb tests to price fire risks; today, every lightbulb carries a safety mark. Nuclear safety works similarly.

We’ve been learning by doing.. In the last 6 months, we’ve audited and insured frontier agent builders like Cursor, ElevenLabs, Lovable and Harvey. For ElevenLabs to get the world’s first agent insurance coverage, a Lloyd’s of London insurer required quarterly technical audits by an auditor of the insurer’s choice. We’ve built infrastructure for quarterly technical audits of any agent, working alongside firms like Gray Swan (technical evaluations) and Schellman (audit).

In the coming weeks, we’ll share our vision for how an insurance ecosystem can work for the frontier, and open source the world’s first actuarial model for frontier AI catastrophic risk.

We’re hiring across all roles in San Francisco. Come do your life’s work with us.

Rune & Rajiv

💬2🔄1❤️5👀1341📊3

What can games like Diplomacy and 1830: The Game of Railroads and Robber Barons teach models about s...

Latent.Space (@latentspacepod) — Tue, 15 Sep 2026 20:19:28 GMT

What can games like Diplomacy and 1830: The Game of Railroads and Robber Barons teach models about strategy and other work-related skills? We spoke to @goodstartlabs CEO @alxai_ (an alumni of @every) to find out. latent.space/p/good-start-l…

💬3🔄4❤️11👀1976📊7

full transcript and writeup: https://t.co/HI38P6XXrF

Latent.Space (@latentspacepod) — Tue, 15 Sep 2026 19:04:21 GMT

full transcript and writeup: x.com/latentspacepod…

💬0🔄0❤️0👀532

🆕 Humanity’s Last Invention — with @RichardSocher! https://t.co/ZSeHSpcC1g @recursive_si is the l...

Latent.Space (@latentspacepod) — Tue, 15 Sep 2026 19:03:48 GMT

🆕 Humanity’s Last Invention — with @RichardSocher!
youtube.com/watch?v=eDFXtS…g

@recursive_si is the latest neolab to burst on the scene with a $5B fundraise and one of the most impressive cofounder lists ever assembled to tackle open ended, self-improving AI for AI research. We dive into the Eureka Machine, Richard's 10 Spaces of Intelligence (so much left until AGI!), DecaNLP vs @alecrad and why AI peer review is broken, and why this may just be the last invention that humanity has to do on our own.

Timestamps
00:00 The Eureka Machine and Superintelligence
02:23 AI Optimism, Slow Takeoff, and Regulation
07:56 AI Safety, Reward Hacking, and Anthropic’s Constitution
11:49 Alignment, Personalization, and Open Source AI
15:46 Why Richard Started Recursive
20:03 Recursive Self-Improvement and the Founding Team
22:55 Are Today’s LLMs Enough?
29:03 DecaNLP, GPT, and the Rejected Idea Ahead of Its Time
34:38 Open-Endedness and Evolutionary AI
36:38 What Happens When AI Chooses Its Own Goals?
41:16 Superintelligence for Science
42:40 GPUs, Compute, and the Limits of AI Takeoff
45:07 Recursive’s Results: AI Beating Humans and Their Agents
49:14 Reward Engineering and Auto Research
53:12 The AI Economist and Simulating Entire Economies
58:07 LLM Simulations, Personas, and Mode Collapse
03:38 Recursive’s Roadmap, Agents, Search, and Finance
09:13 The Upper Bounds and Spaces of Intelligence
30:21 Goals, High Agency, and Advice for Builders

💬11🔄5❤️18👀12520📊18

soon on LS: - Why you are still underrating Prompting - Claude Code Mods (released yesterday @bcher...

Latent.Space (@latentspacepod) — Tue, 15 Sep 2026 16:20:24 GMT

soon on LS:

- Why you are still underrating Prompting
- Claude Code Mods (released yesterday @bcherny)
- Pacing the Frontier

subscribe: youtube.com/@aidotengineer

Thariq@trq212

just finished recording on latent space

I’m excited about this one- we get very technical about things we haven’t really talked about much yet

💬4🔄1❤️16👀2402📊5

Before co-founding @kepler_ai_hq, @VinooGanesh led Spark at Palantir and built Project Frontline — a...

Latent.Space (@latentspacepod) — Sat, 12 Sep 2026 16:05:09 GMT

Before co-founding @kepler_ai_hq, @VinooGanesh led Spark at Palantir and built Project Frontline — a pioneering program for Forward Deployed Engineers. He takes us through the best practices of FDEs. latent.space/p/forward-depl…

💬6🔄6❤️10👀5806📊12

🆕 project: our Frontier AEO tracker What does Astra prioritize? what sources does it use? What shi...

Latent.Space (@latentspacepod) — Mon, 07 Sep 2026 21:33:53 GMT

🆕 project: our Frontier AEO tracker

What does Astra prioritize? what sources does it use? What shifted from Sol to Astra, and Opus to Fable?
latent.space/p/aeoJ

all our answers within!

💬12🔄5❤️18👀18193📊20

for the first time in a year, @OpenAI's model launch has been better received than a Claude launch. ...

Latent.Space (@latentspacepod) — Fri, 04 Sep 2026 06:13:27 GMT

for the first time in a year, @OpenAI's model launch has been better received than a Claude launch.

latent.space/p/ainews-gpt-6…

this is a huge upset that we had previously not thought possible this year.

swyx@swyx

sorry for the radio silence folks - got sucked into extreme LLM psychosis. but i can confidently say we have crossed over into a new age of AI Engineering and we are never, ever, looking back. 

this isnt even EVERYTHING i did with Astra but i'll append more reports as I publish them on LS!

💬14🔄12❤️128👀25269📊31

We spent >20B tokens throwing @openai's Astra at every AI Engineering task we could think of, beyond...

Latent.Space (@latentspacepod) — Thu, 03 Sep 2026 21:12:41 GMT

We spent >20B tokens throwing @openai's Astra at every AI Engineering task we could think of, beyond cute Blender demos and fun games.

latent.space/p/astra

Here's everything Astra can do, and do so at <$6 an hour (serious):

- choose and train models
- label data (both helping you label and then using your labels for active learning)
- keep pipelines saturated
- instrument and read logs
- deploy and debug entire systems in one shot
- fan out and command and eval subagents (including agents running other models)
- keep coherence over billions of tokens of a single agent thread.

more to come on @swyx's coverage of the Fable- and Astra-class of 2026!

💬25🔄23❤️300👀279996📊140

GitHub invented pull requests, and for 18 years they've been open by default. But now some of the to...

Latent.Space (@latentspacepod) — Tue, 01 Sep 2026 16:26:37 GMT

GitHub invented pull requests, and for 18 years they've been open by default. But now some of the top AI-native open source projects are shutting PRs off, because they've found a better way. latent.space/p/pr-not-welco…

💬6🔄3❤️8👀2590📊10

Lovable is a $13B company with a $500 million annualized revenue run rate. It's just 3 years old, bu...

Latent.Space (@latentspacepod) — Wed, 26 Aug 2026 16:46:10 GMT

Lovable is a $13B company with a $500 million annualized revenue run rate. It's just 3 years old, but essentially competes with Vercel & Cloudflare now — amazing story. And now @Lovable is branching out from AI-powered web app creation and into MCP-powered ‘capabilities’.

We talk to CTO @FabianHedin about the future of SaaS and how Lovable is building a "company brain." latent.space/p/lovable-futu…

💬3🔄0❤️6👀8595📊6

🆕 We’re launching a new Forward Deployed Engineering pod with @realbasilchatha! https://t.co/f72We...

Latent.Space (@latentspacepod) — Tue, 25 Aug 2026 22:33:48 GMT

🆕 We’re launching a new Forward Deployed Engineering pod with @realbasilchatha!
youtube.com/watch?v=MwNvow…s

In just the last few months, $9B+ has been committed by OpenAI, Anthropic, Google, and Microsoft to deploy AI into the enterprise. Model capabilities are there, but it takes a lot to make them reliable and usable for customers at scale.

In our pilot episode, we dive deep into the practical realities of building enterprise voice agents with @decagon, @vapi, @retell, @smallest_ai, and @daily.

We talk about:
-> Why STT -> LLM -> TTS pipelines are used instead of voice-to-voice models
-> How to balance intelligence vs latency for different use cases
-> Why turn-taking is still not a solved problem

Check it out!

💬4🔄1❤️24👀5107📊11

Models keep absorbing the harness into their weights — soon, it will be a harness for human attentio...

Latent.Space (@latentspacepod) — Sat, 22 Aug 2026 10:03:36 GMT

Models keep absorbing the harness into their weights — soon, it will be a harness for human attention rather than for the model. 

A thoughtful analysis of the past and future of agent harnesses by @daniel_mac8latent.space/p/attention-in…

💬6🔄8❤️43👀4058📊20

You are not sufficiently appreciating the scope of RSI. Here's a brief prehistory. https://t.co/f1...

Latent.Space (@latentspacepod) — Sat, 22 Aug 2026 08:14:27 GMT

You are not sufficiently appreciating the scope of RSI. 

Here's a brief prehistory. x.com/i/article/2091…

💬12🔄6❤️99👀13357📊63

Simulating Humanity: 85% accurate digital twins, behavioral foundation models, social physics, & 8 b...

Latent.Space (@latentspacepod) — Fri, 21 Aug 2026 23:41:24 GMT

Simulating Humanity: 85% accurate digital twins, behavioral foundation models, social physics, & 8 billion agents latent.space/p/simile@simile_ai CEO @joon_s_pk explains how AI can move from predicting what people will do to simulating how to shape outcomes, why today’s frontier models still miss how humans actually behave, how digital twins reproduced people with 85% accuracy, why human biases and mistakes have to be learned rather than optimized away, and what it would take to eventually simulate all 8 billion people on Earth.

Your browser does not support the video tag.

💬6🔄5❤️97👀72394📊43

We chat to @mattpocockuk about his /wayfinder skill, which he designed for the "fog of war" — when y...

Latent.Space (@latentspacepod) — Thu, 20 Aug 2026 21:06:34 GMT

We chat to @mattpocockuk about his /wayfinder skill, which he designed for the "fog of war" — when you need to figure out a project but the end state isn’t entirely clear. This is the first in a series of skills we'll be exploring in the coming weeks. latent.space/p/wayfinder-sk…

💬4🔄12❤️142👀102641📊71

@aiDotEngineer https://t.co/sK4MkyF6x0 also the talk from @raymondmfeng is https://t.co/nJX5fHVlTX

Latent.Space (@latentspacepod) — Wed, 19 Aug 2026 19:20:24 GMT

@aiDotEngineerlatent.space/p/community

also the talk from @raymondmfeng is youtube.com/watch?v=k35LeK…

💬0🔄0❤️0👀687

another LS member saw the @aidotengineer posttraining talks on OPSD and is doing a deep dive on toda...

Latent.Space (@latentspacepod) — Wed, 19 Aug 2026 19:19:45 GMT

another LS member saw the @aidotengineer posttraining talks on OPSD and is doing a deep dive on today's paper club! join in (link below)

swyx@swyx

Trajectory have generally impressed me with their tasteful execution on ambitious goals. on our Continual Learning track, @rronak_ gave a very thoughtful overview on how they're tackling the main data problems left in CL, including why GRPO isn't enough and they had to go on-policy.... and then subsequently fix all the issues that come up with it

nice overview from one of the early leaders in this field!

(see the rest of the track for more, this one was quite stacked)