Jacob Effron

Managing Director @redpoint supporting @AbridgeHQ @wearelegora @tryaugie @tryramp @getgarner @AcuityMD @scribehow / AI pod: Unsupervised Learning

[Jacob Effron @jacobeffron]

New Unsupervised Learning with Google AI researchers @NoamShazeer and @jack_w_rae on:

- Scaling test-time compute
- The power of the Mom eval
- The pace of Open Source / DeepSeek
- How AI research is where chemistry was in the 15th century
- Reactions to Ilya on how far test-time compute gets us and Yann LeCun on the limits of models today
- General vs. specialized models
- Implications of AGI and risks

YouTube: https://t.co/M1HsHUfKk3
Spotify: https://t.co/YrMpeAqjsf
Apple: https://t.co/NRZlluiXI0

[Jacob Effron @jacobeffron]

It was a ton of fun to sit down with @SchmidhuberAI who’s been cited as the father of AI by the New York Times and Forbes. We discussed everything going on in the AI ecosystem including:

- Why he thinks RSI work won’t provide the model labs a moat
- What he thinks researchers should be focused on
- Why he thinks data centers will lose massive amounts of value
- Why he’s less worried about AI safety
- What improvements are needed for models

0:00 Intro
1:24 How Close Is Superhuman AI?
2:27 Why ChatGPT Didn't Surprise Him
3:21 The Path to Recursive Self-Improvement
9:01 Will AI Takeoff Feel Sudden?
11:02 Intelligence Means Efficiency
12:32 Advice for Labs: Beyond Human-Biased Data
17:10 Artificial Curiosity and the Theory of Fun
21:33 When Do We Get the AI Scientist?
24:07 AI Chemistry, MOFs, and Carbon Capture
25:04 Robotics Reality Check
28:23 The Data Center Bet: Overbuilt?
34:25 Does Being First to RSI Create a Moat?
38:06 AI Safety and Alignment Skepticism
43:44 Quickfire

YouTube: https://t.co/ITzCeL610T
Apple: https://t.co/YZiUvX3JAW
Spotify: https://t.co/QVlXuAsyVd

[Zhengyang Geng @ZhengyangGeng]

This is an interview I found very interesting from @lukaszkaiser

[Nikhil Krishnan @nikillinit]

A big debate we've been having is how do you make your org "AI-native"? There's no correct answer to this an all of them have tradeoffs.

- You can try to measure inputs, things like token usage. But as we've now seen, people will purposefully use tokens inefficiently to hit arbitrary metrics.

- You can try to measure outputs, things like increase in task completion rate. But how do you assess for task complexity + prevent people from optimizing for taking credit?

- Or you can select for a certain type of willingness to use AI. When Jacob Effron talked to Chris Altchek he mentioned making the entire engineer/product team re-interview for their jobs and show how they're using coding agents, and they plan to do that for the rest of the teams now.

This topic has started a lot of debates in the group chat. What are we even measuring or caring about when it comes to making employees AI-native?

[Jacob Effron @jacobeffron]

@reneeshah123 @sailresearchco 🚀

[Jacob Effron @jacobeffron]

Fresh off their $100M Series C, @nikillinit and I sat down with Cadence CEO @caltchek to discuss changing care models with AI, building voice agents, why hospital at home hasn't scaled as much as people might have expected and a bunch more. Really fun conversation check it out:

Spotify: https://t.co/GUh66TiX8h
Apple: https://t.co/UpUYIhKMok

[Jacob Effron @jacobeffron]

.@_RobToews , on why Google's structural advantages survive falling behind on coding:

[Jacob Effron @jacobeffron]

I sat down with @arimorcos and @_RobToews for our recurring AI vibe check. We got into the tenuous future of open source models, the latest in the lab wars and some really interesting future implications of the compute crunch. It's been far too long since we did one of these and I always have fun jamming with these two. This time we hit on:

▪️ Why near-frontier open weight AI may be disappearing
▪️ Will compute constraints push labs to suspend their own APIs
▪️ Frustrations around Fable
▪️ OpenAI’s future
▪️ Where we are with Recursive Self Improvement and its implications

We also hit on Cursor/xAI, ASML competitors and a ton more.

0:00 Intro
1:40 Coding Agents Cross a Threshold
3:29 Is Open-Weight AI in Retreat?
7:37 Cost Crunch & Scaffolding
12:13 The "Apps Are Cooked" Debate
16:37 Sam Altman Under Scrutiny
19:44 Anthropic's Fable Backlash
23:24 How Big a Step Change Is Fable?
26:50 What's Going On at Google?
33:20 Could the APIs Go Away?
34:11 Breaking the Semiconductor Bottleneck
35:42 Beyond EUV: Atom & X-Ray Lithography
37:23 Implications of a Compute Shortage
40:20 Do Alt Chips Actually Help?
43:43 SpaceX, xAI & the Cursor Acquisition
48:50 How Close Are We to RSI?
52:21 Quickfire

YouTube: https://t.co/Y1gLQ42INw
Spotify: https://t.co/NIKvxtaQq7
Apple: https://t.co/9jx67XidOJ

[Jacob Effron @jacobeffron]

. @arimorcos on the implications of the compute crunch:

[Jacob Effron @jacobeffron]

I sat down with @arimorcos and @_RobToews for our recurring AI vibe check. We got into the tenuous future of open source models, the latest in the lab wars and some really interesting future implications of the compute crunch. It's been far too long since we did one of these and I always have fun jamming with these two. This time we hit on:

▪️ Why near-frontier open weight AI may be disappearing
▪️ Will compute constraints push labs to suspend their own APIs
▪️ Frustrations around Fable
▪️ OpenAI’s future
▪️ Where we are with Recursive Self Improvement and its implications

We also hit on Cursor/xAI, ASML competitors and a ton more.

0:00 Intro
1:40 Coding Agents Cross a Threshold
3:29 Is Open-Weight AI in Retreat?
7:37 Cost Crunch & Scaffolding
12:13 The "Apps Are Cooked" Debate
16:37 Sam Altman Under Scrutiny
19:44 Anthropic's Fable Backlash
23:24 How Big a Step Change Is Fable?
26:50 What's Going On at Google?
33:20 Could the APIs Go Away?
34:11 Breaking the Semiconductor Bottleneck
35:42 Beyond EUV: Atom & X-Ray Lithography
37:23 Implications of a Compute Shortage
40:20 Do Alt Chips Actually Help?
43:43 SpaceX, xAI & the Cursor Acquisition
48:50 How Close Are We to RSI?
52:21 Quickfire

YouTube: https://t.co/Y1gLQ42INw
Spotify: https://t.co/NIKvxtaQq7
Apple: https://t.co/9jx67XidOJ

[Jacob Effron @jacobeffron]

.@arimorcos on why the retreat of open models pushes companies to build their own:

[Jacob Effron @jacobeffron]

I sat down with @arimorcos and @_RobToews for our recurring AI vibe check. We got into the tenuous future of open source models, the latest in the lab wars and some really interesting future implications of the compute crunch. It's been far too long since we did one of these and I always have fun jamming with these two. This time we hit on:

▪️ Why near-frontier open weight AI may be disappearing
▪️ Will compute constraints push labs to suspend their own APIs
▪️ Frustrations around Fable
▪️ OpenAI’s future
▪️ Where we are with Recursive Self Improvement and its implications

We also hit on Cursor/xAI, ASML competitors and a ton more.

0:00 Intro
1:40 Coding Agents Cross a Threshold
3:29 Is Open-Weight AI in Retreat?
7:37 Cost Crunch & Scaffolding
12:13 The "Apps Are Cooked" Debate
16:37 Sam Altman Under Scrutiny
19:44 Anthropic's Fable Backlash
23:24 How Big a Step Change Is Fable?
26:50 What's Going On at Google?
33:20 Could the APIs Go Away?
34:11 Breaking the Semiconductor Bottleneck
35:42 Beyond EUV: Atom & X-Ray Lithography
37:23 Implications of a Compute Shortage
40:20 Do Alt Chips Actually Help?
43:43 SpaceX, xAI & the Cursor Acquisition
48:50 How Close Are We to RSI?
52:21 Quickfire

YouTube: https://t.co/Y1gLQ42INw
Spotify: https://t.co/NIKvxtaQq7
Apple: https://t.co/9jx67XidOJ

[DatologyAI @datologyai]

A spicy take from @arimorcos on @jacobeffron's Unsupervised Learning: frontier APIs may not always be there. The teams that can build their own models won't be exposed when that happens.

[Jacob Effron @jacobeffron]

. @arimorcos on why Frontier Lab APIs might disappear:

[Jacob Effron @jacobeffron]

I sat down with @arimorcos and @_RobToews for our recurring AI vibe check. We got into the tenuous future of open source models, the latest in the lab wars and some really interesting future implications of the compute crunch. It's been far too long since we did one of these and I always have fun jamming with these two. This time we hit on:

▪️ Why near-frontier open weight AI may be disappearing
▪️ Will compute constraints push labs to suspend their own APIs
▪️ Frustrations around Fable
▪️ OpenAI’s future
▪️ Where we are with Recursive Self Improvement and its implications

We also hit on Cursor/xAI, ASML competitors and a ton more.

0:00 Intro
1:40 Coding Agents Cross a Threshold
3:29 Is Open-Weight AI in Retreat?
7:37 Cost Crunch & Scaffolding
12:13 The "Apps Are Cooked" Debate
16:37 Sam Altman Under Scrutiny
19:44 Anthropic's Fable Backlash
23:24 How Big a Step Change Is Fable?
26:50 What's Going On at Google?
33:20 Could the APIs Go Away?
34:11 Breaking the Semiconductor Bottleneck
35:42 Beyond EUV: Atom & X-Ray Lithography
37:23 Implications of a Compute Shortage
40:20 Do Alt Chips Actually Help?
43:43 SpaceX, xAI & the Cursor Acquisition
48:50 How Close Are We to RSI?
52:21 Quickfire

YouTube: https://t.co/Y1gLQ42INw
Spotify: https://t.co/NIKvxtaQq7
Apple: https://t.co/9jx67XidOJ

[Jacob Effron @jacobeffron]

I sat down with @arimorcos and @_RobToews for our recurring AI vibe check. We got into the tenuous future of open source models, the latest in the lab wars and some really interesting future implications of the compute crunch. It's been far too long since we did one of these and I always have fun jamming with these two. This time we hit on:

▪️ Why near-frontier open weight AI may be disappearing
▪️ Will compute constraints push labs to suspend their own APIs
▪️ Frustrations around Fable
▪️ OpenAI’s future
▪️ Where we are with Recursive Self Improvement and its implications

We also hit on Cursor/xAI, ASML competitors and a ton more.

0:00 Intro
1:40 Coding Agents Cross a Threshold
3:29 Is Open-Weight AI in Retreat?
7:37 Cost Crunch & Scaffolding
12:13 The "Apps Are Cooked" Debate
16:37 Sam Altman Under Scrutiny
19:44 Anthropic's Fable Backlash
23:24 How Big a Step Change Is Fable?
26:50 What's Going On at Google?
33:20 Could the APIs Go Away?
34:11 Breaking the Semiconductor Bottleneck
35:42 Beyond EUV: Atom & X-Ray Lithography
37:23 Implications of a Compute Shortage
40:20 Do Alt Chips Actually Help?
43:43 SpaceX, xAI & the Cursor Acquisition
48:50 How Close Are We to RSI?
52:21 Quickfire

YouTube: https://t.co/Y1gLQ42INw
Spotify: https://t.co/NIKvxtaQq7
Apple: https://t.co/9jx67XidOJ

[Abridge @AbridgeHQ]

@WSJ Exclusive: @nvidia + Abridge on the first clinical conversation foundation model

"Generic models are powerful, but clinical intelligence—it still has to be trained, it has to be shaped, and it has to be evaluated against real-world conditions." —@ShivdevRao, Abridge CEO and Co-Founder

So many more announcements to come at today’s Abridge Keynote: https://t.co/OseKPvPhed

It’s all coming together. 12PM EDT

[Jacob Effron @jacobeffron]

Transformer co-author, @lukaszkaiser, on if big labs lose the ability to make concentrated bets as they scale to thousands of employees:

[Jacob Effron @jacobeffron]

I sat down with @lukaszkaiser to get into whether the architecture he helped invent is actually enough, and what's next in generalization, coding agents, RL and more. Lukasz co-authored "Attention Is All You Need," the paper that introduced the transformer and worked on reasoning models at OpenAI so he’s been a key part of major shifts in the field. We hit on:

▪️ The case for and against a new architecture coming after the transformer
▪️ What’s required for model generalization in the physical world
▪️ How much coding agents have improved his AI research productivity
▪️ The next domains for RL
▪️ Why Anthropic initially won coding
▪️ Future research directions he’s excited about

0:00 Intro
1:12 Transformers vs. Human Learning
8:37 How Do We Get Physical World Generalization?
10:52 What Comes After Transformers
13:59 How Much Have Agents Improved Lukasz's AI Research Productivity?
17:21 How Close Is an AI Research Intern?
26:06 RL Beyond Verifiable Tasks
35:38 App Companies: Build Models or Lean on Labs?
46:21 Multimodal Is Still Missing Something
49:46 OpenAI's Bet on Reasoning
55:26 The AI Coding Wars
59:26 Focus vs. Keeping Embers Burning
1:02:09 Open Source vs. Closed Source Gap
1:05:15 Quickfire

YouTube: https://t.co/wetErBD4B8
Spotify: https://t.co/FSYWR39ep2
Apple: https://t.co/j2omVI8xEa

[Jacob Effron @jacobeffron]

. @lukaszkaiser on hardware progress: the GPU he has under his desk today is more powerful than the entire cluster they used to train the original Transformer:

[Jacob Effron @jacobeffron]

I sat down with @lukaszkaiser to get into whether the architecture he helped invent is actually enough, and what's next in generalization, coding agents, RL and more. Lukasz co-authored "Attention Is All You Need," the paper that introduced the transformer and worked on reasoning models at OpenAI so he’s been a key part of major shifts in the field. We hit on:

▪️ The case for and against a new architecture coming after the transformer
▪️ What’s required for model generalization in the physical world
▪️ How much coding agents have improved his AI research productivity
▪️ The next domains for RL
▪️ Why Anthropic initially won coding
▪️ Future research directions he’s excited about

0:00 Intro
1:12 Transformers vs. Human Learning
8:37 How Do We Get Physical World Generalization?
10:52 What Comes After Transformers
13:59 How Much Have Agents Improved Lukasz's AI Research Productivity?
17:21 How Close Is an AI Research Intern?
26:06 RL Beyond Verifiable Tasks
35:38 App Companies: Build Models or Lean on Labs?
46:21 Multimodal Is Still Missing Something
49:46 OpenAI's Bet on Reasoning
55:26 The AI Coding Wars
59:26 Focus vs. Keeping Embers Burning
1:02:09 Open Source vs. Closed Source Gap
1:05:15 Quickfire

YouTube: https://t.co/wetErBD4B8
Spotify: https://t.co/FSYWR39ep2
Apple: https://t.co/j2omVI8xEa

[Jacob Effron @jacobeffron]

Transformer co-author, @lukaszkaiser, on why he prefers Codex:

[Jacob Effron @jacobeffron]

I sat down with @lukaszkaiser to get into whether the architecture he helped invent is actually enough, and what's next in generalization, coding agents, RL and more. Lukasz co-authored "Attention Is All You Need," the paper that introduced the transformer and worked on reasoning models at OpenAI so he’s been a key part of major shifts in the field. We hit on:

▪️ The case for and against a new architecture coming after the transformer
▪️ What’s required for model generalization in the physical world
▪️ How much coding agents have improved his AI research productivity
▪️ The next domains for RL
▪️ Why Anthropic initially won coding
▪️ Future research directions he’s excited about

0:00 Intro
1:12 Transformers vs. Human Learning
8:37 How Do We Get Physical World Generalization?
10:52 What Comes After Transformers
13:59 How Much Have Agents Improved Lukasz's AI Research Productivity?
17:21 How Close Is an AI Research Intern?
26:06 RL Beyond Verifiable Tasks
35:38 App Companies: Build Models or Lean on Labs?
46:21 Multimodal Is Still Missing Something
49:46 OpenAI's Bet on Reasoning
55:26 The AI Coding Wars
59:26 Focus vs. Keeping Embers Burning
1:02:09 Open Source vs. Closed Source Gap
1:05:15 Quickfire

YouTube: https://t.co/wetErBD4B8
Spotify: https://t.co/FSYWR39ep2
Apple: https://t.co/j2omVI8xEa

[Jacob Effron @jacobeffron]

Transformer co-author, @lukazkaiser, talks about how much agents increased his productivity in AI research::

[Jacob Effron @jacobeffron]

I sat down with @lukaszkaiser to get into whether the architecture he helped invent is actually enough, and what's next in generalization, coding agents, RL and more. Lukasz co-authored "Attention Is All You Need," the paper that introduced the transformer and worked on reasoning models at OpenAI so he’s been a key part of major shifts in the field. We hit on:

▪️ The case for and against a new architecture coming after the transformer
▪️ What’s required for model generalization in the physical world
▪️ How much coding agents have improved his AI research productivity
▪️ The next domains for RL
▪️ Why Anthropic initially won coding
▪️ Future research directions he’s excited about

0:00 Intro
1:12 Transformers vs. Human Learning
8:37 How Do We Get Physical World Generalization?
10:52 What Comes After Transformers
13:59 How Much Have Agents Improved Lukasz's AI Research Productivity?
17:21 How Close Is an AI Research Intern?
26:06 RL Beyond Verifiable Tasks
35:38 App Companies: Build Models or Lean on Labs?
46:21 Multimodal Is Still Missing Something
49:46 OpenAI's Bet on Reasoning
55:26 The AI Coding Wars
59:26 Focus vs. Keeping Embers Burning
1:02:09 Open Source vs. Closed Source Gap
1:05:15 Quickfire

YouTube: https://t.co/wetErBD4B8
Spotify: https://t.co/FSYWR39ep2
Apple: https://t.co/j2omVI8xEa

[Xander Dunn @xanderai]

tl;dr It’s the Age of Research

Something is off about models:
1) Training is so sample inefficient
2) Very long thinking trajectories: models will do the right thing only after exhausting all other possibilities
3) Generalization is whack. Waymo can’t handle construction on highway. No teenager would have this problem.

Unclear why December was an inflection point for coding agents. No single clear attributable change.
Google is still pre-December on coding capabilities.

5-10x speed up in work. 3 weeks to implement paper in ye olden days. 2 days with codex + can do many things in parallel.

Humans see, hear, talk, everything all at once. Of course that’s how it should be for models.

Big update on how quickly we got to an intern level coding agent. Didn’t expect that in 2025.

[Jacob Effron @jacobeffron]

@lukaszkaiser That was a ton of fun!

[Jacob Effron @jacobeffron]

.@lukaszkaiser on why Anthropic was first to break through in coding:

[Jacob Effron @jacobeffron]

I sat down with @lukaszkaiser to get into whether the architecture he helped invent is actually enough, and what's next in generalization, coding agents, RL and more. Lukasz co-authored "Attention Is All You Need," the paper that introduced the transformer and worked on reasoning models at OpenAI so he’s been a key part of major shifts in the field. We hit on:

▪️ The case for and against a new architecture coming after the transformer
▪️ What’s required for model generalization in the physical world
▪️ How much coding agents have improved his AI research productivity
▪️ The next domains for RL
▪️ Why Anthropic initially won coding
▪️ Future research directions he’s excited about

0:00 Intro
1:12 Transformers vs. Human Learning
8:37 How Do We Get Physical World Generalization?
10:52 What Comes After Transformers
13:59 How Much Have Agents Improved Lukasz's AI Research Productivity?
17:21 How Close Is an AI Research Intern?
26:06 RL Beyond Verifiable Tasks
35:38 App Companies: Build Models or Lean on Labs?
46:21 Multimodal Is Still Missing Something
49:46 OpenAI's Bet on Reasoning
55:26 The AI Coding Wars
59:26 Focus vs. Keeping Embers Burning
1:02:09 Open Source vs. Closed Source Gap
1:05:15 Quickfire

YouTube: https://t.co/wetErBD4B8
Spotify: https://t.co/FSYWR39ep2
Apple: https://t.co/j2omVI8xEa