Uploaded November 2026 | Updated November 2026, 3 days ago
Why do PDFs still break AI? Reducto co-founder and CEO Adit Abraham joins Allen Park on Latent Space Cooking to talk AI document parsing and make a steak sandwich with burrata, torched peaches, and chimichurri.
From the YC pivot out of long-term memory to Reducto R1 and Deep Extract, Adit explains why even frontier models can miss rows in long documents, how his team catches failures in production, and what it takes to go from reading a PDF to reliably editing one. He also revisits the viral steak tweet that brought him into the kitchen.
Sponsorship and business inquiries: business@latent.space
We talk about:
• Why complex tables, handwriting, and unusual layouts still trip up AI
• When specialized models and extraction pipelines beat a general-purpose model
• Why 99% accuracy can still fall short, and where citations and human review help
• What independent benchmarks and production evals reveal about messy documents
• What changes at a billion pages a month, the scale Adit shared at filming
• What Reducto looks for in engineers, and why it lets candidates use AI in interviews
• What a Claude Code moment for office work might look like
00:00 Cold Open
00:13 Meet Adit Abraham and the Viral Steak Tweet
00:57 From Remembrall to Reducto: YC and the Pivot
01:57 Why PDFs Are Still Hard for AI
03:35 What AI Agents Need to Do With Documents
04:19 Model Routing for Tricky Pages and Tables
05:29 Build vs. Buy: When to Train Your Own Models
07:00 Deep Extract, Frontier Models, and Missing Rows
08:20 Why Everyone Wins Their Own Benchmark
09:29 Reducto R1: Accuracy, Cost, and a Billion Pages a Month
10:57 When 99% Accuracy Isn’t Enough
13:05 Continuous Evals and Production Failures
16:49 How Reducto Uses AI Agents Internally
18:28 Hiring With AI and the Eight-Hour Onsite
20:17 Early Hires and What Obsession Looks Like
23:17 Blowtorching Peaches
24:24 Scaling From a Small Team to Series B
27:34 Competition and the Data Flywheel
30:52 What Better Parsing Unlocks for Customers
32:11 Finding a Cofounder at MIT
34:34 The Anthropic Hackathon, Long-Term Memory, and RAG
38:49 From Reading PDFs to Redlining Contracts
43:19 A Claude Code Moment for Office Work
46:13 What’s Next for Reducto: Extraction, Citations, and Cost
47:51 Assembling the Steak Sandwich
49:55 The Steak Allegations and Taste Test
Why do PDFs still break AI? Reducto co-founder and CEO Adit Abraham joins Allen Park on Latent Space Cooking to talk AI document parsing and make a steak sandwich with burrata, torched peaches, and chimichurri.
From the YC pivot out of long-term memory to Reducto R1 and Deep Extract, Adit explains why even frontier models can miss rows in long documents, how his team catches failures in production, and what it takes to go from reading a PDF to reliably editing one. He also revisits the viral steak tweet that brought him into the kitchen.
Sponsorship and business inquiries: business@latent.space
We talk about:
• Why complex tables, handwriting, and unusual layouts still trip up AI
• When specialized models and extraction pipelines beat a general-purpose model
• Why 99% accuracy can still fall short, and where citations and human review help
• What independent benchmarks and production evals reveal about messy documents
• What changes at a billion pages a month, the scale Adit shared at filming
• What Reducto looks for in engineers, and why it lets candidates use AI in interviews
• What a Claude Code moment for office work might look like
00:00 Cold Open
00:13 Meet Adit Abraham and the Viral Steak Tweet
00:57 From Remembrall to Reducto: YC and the Pivot
01:57 Why PDFs Are Still Hard for AI
03:35 What AI Agents Need to Do With Documents
04:19 Model Routing for Tricky Pages and Tables
05:29 Build vs. Buy: When to Train Your Own Models
07:00 Deep Extract, Frontier Models, and Missing Rows
08:20 Why Everyone Wins Their Own Benchmark
09:29 Reducto R1: Accuracy, Cost, and a Billion Pages a Month
10:57 When 99% Accuracy Isn’t Enough
13:05 Continuous Evals and Production Failures
16:49 How Reducto Uses AI Agents Internally
18:28 Hiring With AI and the Eight-Hour Onsite
20:17 Early Hires and What Obsession Looks Like
23:17 Blowtorching Peaches
24:24 Scaling From a Small Team to Series B
27:34 Competition and the Data Flywheel
30:52 What Better Parsing Unlocks for Customers
32:11 Finding a Cofounder at MIT
34:34 The Anthropic Hackathon, Long-Term Memory, and RAG
38:49 From Reading PDFs to Redlining Contracts
43:19 A Claude Code Moment for Office Work
46:13 What’s Next for Reducto: Extraction, Citations, and Cost
47:51 Assembling the Steak Sandwich
49:55 The Steak Allegations and Taste Test





![[State of AI Startups] Memory/Learning, RL Envs & DBT-Fivetran — Sarah Catanzaro, Amplify
From investing through the modern data stack era (DBT, Fivetran, and the analytics explosion) to now investing at the frontier of AI infrastructure and applications at *Amplify Partners,* *Sarah Catanzaro* has spent years at the intersection of data, compute, and intelligence—watching categories emerge, merge, and occasionally disappoint. We caught up with Sarah live at *NeurIPS 2025* to dig into the state of AI startups heading into 2026: why $100M+ seed rounds with no near-term roadmap are now the norm (and why that terrifies her), what the DBT-Fivetran merger really signals about the modern data stack (spoiler: its not dead, just ready for IPO), how frontier labs are using DBT and Fivetran to manage training data and agent analytics at scale, why data catalogs failed as standalone products but might succeed as metadata services for agents, the consumerization of AI and why personalization (memory, continual learning, K-factor) is the 2026 unlock for retention and growth, why she thinks *RL environments are a fad* and real-world logs beat synthetic clones every time, and her thesis for the most exciting AI startups: companies that marry hard research problems (RAG, rule-following, continual learning) with killer applications that were simply impossible before.
We discuss:
* The *DBT-Fivetran merger:* not the death of the modern data stack, but a path to IPO scale (targeting $600M+ combined revenue) and a signal that both companies were already winning their categories
* How *frontier labs use data infrastructure:* DBT and Fivetran for training data curation, agent analytics, and managing increasingly complex interactions—plus the rise of transactional databases (RocksDB) and efficient data loading (Vortex) for GPU-bound workloads
* Why *data catalogs failed:* built for humans when they should have been built for machines, focused on discoverability when the real opportunity was governance, and ultimately subsumed as features inside Snowflake, DBT, and Fivetran
* The *$100M+ seed phenomenon:* raising massive rounds at billion-dollar valuations with no 6-month roadmap, seven-day decision windows, and founders optimizing for signal (were a unicorn) over partnership or dilution discipline
* Why *world models are overhyped but underspecified:* three competing definitions, unclear generalization across use cases (video games ≠ robotics ≠ autonomous driving), and a research problem masquerading as a product category
* The 2026 theme: *consumerization of AI via personalization*—memory management, continual learning, and solving retention/churn by making products learn skills, preferences, and adapt as the world changes (not just storing facts in cursor rules)
* Why *RL environments are a fad:* labs are paying 7–8 figures for synthetic clones when real-world logs, traces, and user activity (à la Cursor) are richer, cheaper, and more generalizable
* Sarahs investment thesis: *research-driven applications* that solve hard technical problems (RAG for Harvey, rule-following for Sierra, continual learning for the next killer app) and unlock experiences that were impossible before
* Infrastructure bets: *memory, continual learning, stateful inference,* and the systems challenges of loading/unloading personalized weights at scale
* Why K-factor and growth fundamentals matter again: AI felt magical in 2023–2024, but as the magic fades, retention and virality are back—and most AI founders have never heard of K-factor
—
Sarah Catanzaro
* X: https://x.com/sarahcat21
* Amplify Partners: https://amplifypartners.com/
Where to find Latent Space
* X: https://x.com/latentspacepod
* Substack: https://www.latent.space/
00:00:00 Introduction: Sarah Catanzaros Journey from Data to AI
00:01:02 The DBT-Fivetran Merger: Not the End of the Modern Data Stack
00:05:26 Data Catalogs and What Went Wrong
00:08:16 Data Infrastructure at AI Labs: Surprising Insights
00:10:13 The Crazy Funding Environment of 2024-2025
00:17:18 World Models: Hype, Confusion, and Market Potential
00:18:59 Memory Management and Continual Learning: The Next Frontier
00:23:27 Agent Environments: Just a Fad?
00:25:48 The Perfect AI Startup: Research Meets Application
00:28:02 Closing Thoughts and Where to Find Sarah [State of AI Startups] Memory/Learning, RL Envs & DBT-Fivetran — Sarah Catanzaro, Amplify](https://i.ytimg.com/vi/LdVldygYE6I/mqdefault.jpg)

 — Performance, pricing and safeguards.
- [The brain and the hands](https://www.anthropic.com/research/managed-agents) — The Managed Agents architecture.
- [How Anthropic contains Claude](https://www.anthropic.com/research/how-we-contain-claude) — Isolation, credentials and network controls.
- [Harnesses for long-running agents](https://www.anthropic.com/research/effective-harnesses-for-long-running-agents) — Maintaining progress across context windows.
Sponsorship and business inquiries: business@latent.space How Anthropic ships the #1 LLM in the World — with Katelyn Lesse and Angela Jiang](https://i.ytimg.com/vi/MRQTlwcW-aQ/mqdefault.jpg)


![[State of Code Evals] After SWE-bench, Code Clash & SOTA Coding Benchmarks recap — John Yang
From creating *SWE-bench* in a Princeton basement to shipping *CodeClash,* *SWE-bench Multimodal,* and *SWE-bench Multilingual,* *John Yang* has spent the last year and a half watching his benchmark become the de facto standard for evaluating AI coding agents—trusted by Cognition (Devin), OpenAI, Anthropic, and every major lab racing to solve software engineering at scale. We caught up with John live at *NeurIPS 2025* to dig into the state of code evals heading into 2026: why *SWE-bench went from ignored (October 2023) to the industry standard* after Devins launch (and how Walden emailed him two weeks before the big reveal), how the benchmark evolved from Django-heavy to *nine languages across 40 repos* (JavaScript, Rust, Java, C, Ruby), why *unit tests as verification are limiting* and long-running agent tournaments might be the future (CodeClash: agents maintain codebases, compete in arenas, and iterate over multiple rounds), the *proliferation of SWE-bench variants* (SWE-bench Pro, SWE-bench Live, SWE-Efficiency, AlgoTune, SciCode) and how benchmark authors are now justifying their splits with curation techniques instead of just more repos, why *Tau-benchs impossible tasks controversy* is actually a feature not a bug (intentionally including impossible tasks flags cheating), the tension between *long autonomy (5-hour runs) vs. interactivity* (Cognitions emphasis on fast back-and-forth), how *Terminal-bench unlocked creativity* by letting PhD students and non-coders design environments beyond GitHub issues and PRs, the *academic data problem* (companies like Cognition and Cursor have rich user interaction data, academics need user simulators or compelling products like LMArena to get similar signal), and his vision for *CodeClash as a testbed for human-AI collaboration*—freeze model capability, vary the collaboration setup (solo agent, multi-agent, human+agent), and measure how interaction patterns change as models climb the ladder from code completion to full codebase reasoning.
We discuss:
* Johns path: *Princeton → SWE-bench (October 2023) → Stanford PhD with Diyi Yang and the Iris Group,* focusing on code evals, human-AI collaboration, and long-running agent benchmarks
* The *SWE-bench origin story:* released October 2023, mostly ignored until *Cognitions Devin launch* kicked off the arms race (Walden emailed John two weeks before: we have a good number)
* *SWE-bench Verified:* the curated, high-quality split that became the standard for serious evals
* *SWE-bench Multimodal and Multilingual:* nine languages (JavaScript, Rust, Java, C, Ruby) across 40 repos, moving beyond the Django-heavy original distribution
* The *SWE-bench Pro controversy:* independent authors used the SWE-bench name without Johns blessing, but hes okay with it (congrats to them, its a great benchmark)
* *CodeClash:* Johns new benchmark for *long-horizon development*—agents maintain their own codebases, edit and improve them each round, then compete in arenas (programming games like Halite, economic tasks like GDP optimization)
* *SWE-Efficiency* (Jeffrey Maugh, Johns high school classmate): optimize code for speed without changing behavior (parallelization, SIMD operations)
* *AlgoTune, SciCode, Terminal-bench, Tau-bench, SecBench, SRE-bench:* the Cambrian explosion of code evals, each diving into different domains (security, SRE, science, user simulation)
* The *Tau-bench impossible tasks debate:* some tasks are underspecified or impossible, but John thinks thats actually a feature (flags cheating if you score above 75%)
* *Cognitions research focus:* codebase understanding (retrieval++), helping humans understand their own codebases, and automatic context engineering for LLMs (research sub-agents)
* The vision: *CodeClash as a testbed for human-AI collaboration*—vary the setup (solo agent, multi-agent, human+agent), freeze model capability, and measure how interaction changes as models improve
—
John Yang
* SWE-bench: https://www.swebench.com
* X: https://x.com/jyangballin
00:00:00 Introduction: John Yang on SWE-bench and Code Evaluations
00:00:31 SWE-bench Origins and Devons Impact on the Coding Agent Arms Race
00:01:09 SWE-bench Ecosystem: Verified, Pro, Multimodal, and Multilingual Variants
00:02:17 Moving Beyond Django: Diversifying Code Evaluation Repositories
00:03:08 Code Clash: Long-Horizon Development Through Programming Tournaments
00:04:41 From Halite to Economic Value: Designing Competitive Coding Arenas
00:06:04 Ofirs Lab: SWE-ficiency, AlgoTune, and SciCode for Scientific Computing
00:07:52 The Benchmark Landscape: TAU-bench, Terminal-bench, and User Simulation
00:09:20 The Impossible Task Debate: Refusals, Ambiguity, and Benchmark Integrity
00:12:32 The Future of Code Evals: Long Autonomy vs Human-AI Collaboration
00:14:37 Call to Action: User Interaction Data and Codebase Understanding Research [State of Code Evals] After SWE-bench, Code Clash & SOTA Coding Benchmarks recap — John Yang](https://i.ytimg.com/vi/MxB-xRGXxkk/mqdefault.jpg)