Episode Summary
Executive Summary: Didi Das discusses the evolution of enterprise AI through Glean, Anthropic, and a portfolio of AI infrastructure/app companies. He argues that enterprise search is hard, defensible work beyond “just adding AI,” Anthropic’s growth reflects genuine product-market fit, and the biggest opportunities lie in harder technical layers like model infrastructure, interpretability, and fast coding agents—while warning that AI may erode engineers’ deep problem-solving skills.
Main Topics: Glean’s enterprise search moat: Didi explains why Glean succeeded by doing the difficult, unglamorous work of building enterprise search properly before AI made the category sexy. He emphasizes deployment complexity, onboarding, ranking, and enterprise data integration as real moats. Competition, APIs, and SaaS rate limits: He pushes back on SaaS vendors restricting API access and argues customers own their data. He also addresses why frontier labs entering enterprise search is less threatening than it looks, given the sales and customization burden. Anthropic’s growth and product strategy: Didi describes Anthropic as a generational company with unusually fast growth, strong retention, and product innovation such as Claude Code. He argues its success comes from talent freedom, model quality, and enterprise traction, not just PR. Model-layer vs app-layer defensibility: The conversation explores whether value accrues more to models or applications. Didi argues the hardest layer in the stack tends to capture value and that model companies may be more defensible than many app layers, especially when they keep innovating. Portfolio bets on AI infrastructure and research: He highlights investments in OpenRouter, Goodfire, Prime Intellect, Whisper, and a stealth diffusion-model company, framing them as bets on pain points or future market structure rather than obvious near-term categories. Human-centered coding agents and AI-assisted development: Didi warns that vibe coding and overly async agents can weaken engineers’ muscles for deep work, and advocates for “fast agents” that assist reading/comprehension while keeping humans in control of writing and judgment. Reflexivity, capital, and AI market dynamics: He discusses how large funding rounds and infrastructure commitments can create self-fulfilling advantages, shaping competition and making it harder for smaller companies to enter markets once capital and momentum accumulate.
Key Arguments: Glean’s moat was not AI hype; it was years of hard enterprise-search work, including ranking, onboarding, integrations, and last-mile customer deployment. Enterprise search is inherently hard because consumer-search signals do not transfer well to enterprise settings: lower query volume, less feedback data, more freshness, and more domain-specific intent. SaaS companies limiting API access often hurts customers more than it protects vendors, because search over their data does not cannibalize the vendor’s core business. Anthropic’s enterprise success is driven by model quality, retention, and product decisions like Claude Code—not just branding or public perception. Frontier labs can enter enterprise search, but the effort required (sales, FDEs, customization, integrations) is large and may not move the needle for billion-dollar revenue labs. The hardest layer in the stack is often where value accrues; model labs may be more defensible than app layers if the underlying problem is more difficult. Research-heavy startups are high risk but can produce outsized outcomes if they follow a credible path from technical insight to future product relevance. Fast coding agents should support human thinking rather than replace it; async agents are better for commoditized work, while high-focus work benefits from human-in-the-loop assistance. Vibe coding can create new security risks because developers may stop reading code carefully, making hidden vulnerabilities harder to detect. Capital deployment and big strategic bets can alter market structure by deterring competition and creating reflexive momentum around a category.
Data Points: Glean valuation: $7 billion - Didi says Glean is now valued around $7B, up from an earlier ~$1B level. Claude launch timing: Claude 1 in March 2023; Claude 2 in July 2023 - Used to anchor how long ago the last appearance felt. Anthropic market share (2023): OpenAI 50%; Anthropic 12% - Enterprise LLM API spend share from Menlo survey data. Anthropic market share (mid-2025): OpenAI 25%; Anthropic 32% - Enterprise LLM API spend share from Menlo survey data. Anthropic revenue scale: North of $1B annual revenue scale - Didi characterizes Anthropic and OpenAI as billion-dollar-revenue-scale companies. Glean revenue scale: Several hundred million dollars - Used as contrast to Anthropic/OpenAI scale. Anthropic retention: 80% one-year retention - Cited from SignalFire-style tracking of employee retention. Anthology Fund size: $100 million - The Menlo/Anthropic ecosystem fund discussed in the episode. Anthology Fund portfolio count: About 40 companies - Didi says the fund has invested in roughly 40 companies. Anthropic spending in one year: $7 billion total; $2 billion inference; $5 billion R&D - Used to frame compute economics and the scale of infrastructure spending. OpenAI weekly active users: 800 million - Referenced when discussing consumer demand and model quality. OpenAI inference spend (as mentioned): $2 billion - Part of the compute breakdown discussed in the episode. OpenAI R&D spend (as mentioned): $5 billion - Part of the compute breakdown discussed in the episode. Whisper zero-edit rate: Over 80% - Didi cites this as a standout metric for the voice dictation product.
Pivotal Quotes: "It’s such a boring, unsexy company that became sexy later." — Didi Das: On how Glean went from an unglamorous category to a high-value enterprise AI company. "The moat is just we did the hard work." — Didi Das: Explaining why Glean’s defensibility comes from execution on enterprise search fundamentals. "Brain surgery for LLMs" — Didi Das: His catchphrase for mechanistic interpretability and why Goodfire matters.
Implications: Enterprise AI winners will likely come from hard technical and distribution work, not just model wrappers. Frontier labs may keep absorbing value, but durable app companies can still win if they solve gnarly, real workflows better than anyone else.
From the Episode
Now, being on venture side, I have a bit of a different take than I would have had at Glean. But broadly, one of the things that I love about Glean is it's such a boring, unsexy company that became sexy later. So, from 2019, I remember going to parties in the Bay Area and I would say enterprise search, and it's shutting down the conversation right there. You know, like nobody would ever ask a counter question if you said enterprise search. They're like, That sounds boring as hell. Leave me alone. Like, and fast forward to 2022. Enterprise search gets more, got more conversations. I was like, interesting. Tell me how you're doing this search. I think what was nice about that observation is in those three years, we did a lot of work and didn't take shortcuts on a lot of things that ended up generating a lot of value for us now. And I can go into what all of those things are, but if you look at Glean from a high-level business, it is top-down enterprise sales. It's very hard to rip and replace. We have, we expand.
Story at all. I really think we did all the hard work to build search, and AI happened to accelerate our go-to-market motion at the right time. And now I see companies trying to tack on search. It's not easy. I know the kind of like last mile stuff we did for some of our customers. And I just know that when I think about other companies, I'm like, would you really go all that distance? It's not a moat. The moat is just we did the hard work. And so I'm pretty happy. I mean, things. Can go any direction, but I'm pretty happy with the way Glean's going right now. And just to spell out the two main challenges: so, one is obviously Claude, I think today launched enterprise search. I was going to say, I have screenshots. Did you see, like, we're introducing enterprise search? I'm like, son of a gun. And then the other side, you have the data providers adding this rate limits, kind of like Salesforce is done with Slack. It feels like that part is more challenging than the competition from other companies. Yeah.
To understand why models say what they'd say. It's brain surgery for LLMs, is my catchphrase. But doesn't apply to LLMs only, all models. And that is a pretty important insight into deploying AI at scale. Yeah. And you don't know the business model yet, don't need to. It's not that much. There are some ideas that we have, but not ready to talk about publicly, and some that are working also. It's not right to be publicly. Does it feel worthwhile to do this on such small models? Because I think most of the work is done on the Open source releases, like, how much of a gap is there between what they're able to do and then translate that into doing it for? Like, they've shown that even for the biggest open source models, you like, even for like DeepSeq's big models, they can do it. And in general, like scaling is not the bottleneck. Obviously, access to the weights would be a bottleneck, but not but they're in the anthology fund, so they can work with anthravic. They can work with anthravic, but they don't have cloud access. Yeah.
About Latent Space: The AI Engineer Podcast
The podcast by and for AI Engineers! In 2025, over 10 million readers and listeners came to Latent Space to hear about news, papers and interviews in Software 3.0. We cover Foundation Models changing every domain in Code Generation, Multimodality, AI Agents, GPU Infra and more, directly from the founders, builders, and thinkers involved in pushing the cutting edge. Striving to give you both the definitive take on the Current Thing down to the first introduction to the tech you'll be using in the next 3 months! We break news and exclusive interviews from OpenAI, Anthropic, Gemini, Meta (Soumith Chintala), Sierra (Bret Taylor), tiny (George Hotz), Databricks/MosaicML (Jon Frankle), Modular (Chris Lattner), Answer.ai (Jeremy Howard), et al. Full show notes always on https://latent.space
View all episodes from Latent Space: The AI Engineer Podcast