Apr 10, 2026 · 58m · mad
Anthropic’s Felix Rieseberg: Claude Cowork, Mythos, and the SaaS Extinction
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of The MAD Podcast, Anthropic Engineering Manager Felix Rieseberg discusses the emergence of frontier models like Claude Mythos, the architecture of Claude Cowork, and how rapid AI advancement is shifting software development from traditional coding to natural language and product user experience.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Matt holds 18.7% of the talking time here. How this is scored →
Matt as informed peer 3.1 Guest teaching 3.3 Guest disagreement 0.7 Matt pushing back 1.9
speaking balance: gold is Matt, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 55:11 Spicy reframe on AI UI conventions
Felix rejects the universal AI design pattern, arguing that developers reflexively add unnecessary chat sidebars to products instead of thinking through actual utility.
Hardest push from Matt ▶ 22:50 Devil's advocate challenge on local vs cloud AIMatt directly challenges Felix's core thesis on local compute, citing Anthropic's Vercept acquisition to ask why cloud-based systems wouldn't render local AI obsolete.
Biggest teaching moment ▶ 11:42 Debunking the 10-day build mythFelix corrects the popular tech lore that Cowork was written from scratch in 10 days, educating the host on how modern software builds on deep underlying internal research and libraries.
Matt holds his own ▶ 44:30 Citing public market SaaS-pocalypse impactMatt demonstrates deep industry insight by highlighting how minor product updates from Anthropic triggered a massive sell-off in public SaaS equities.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Matt as informed peer | Guest teaching | Guest disagreement | Matt pushing back | Why |
|---|---|---|---|---|---|---|
| Unveiling Claude Mythos Preview and Security Capabilities | 2 | 3 | 0 | 1 | Matt introduces Project Glasswing and Claude Mythos Preview, asking Felix to elaborate on his tweet about step function changes. Felix collaboratively explains internal testing and the cybersecurity breakthrough. | |
| Project Glasswing and Securing Infrastructure | 2 | 3 | 0 | 2 | Matt probes whether Mythos will impact Cowork and asks about its closed deployment strategy. Felix explains the rationale behind Project Glasswing to protect core infrastructure before public release. | |
| Model Categorization and Anthropic's Safety Philosophy | 3 | 4 | 1 | 2 | Matt asks if Mythos is part of the Sonnet line and characterizes the moment as terrifying. Felix clarifies model categorization and explains how product feedback loops inform research. | |
| Debunking the Lore: The 10-Day Build of Claude Cowork | 3 | 5 | 2 | 3 | Matt brings up the industry lore that Cowork was built in 10 days. Felix gently reframes this narrative, clarifying that while his team sprinted for 10 days, software builds on existing internal research and libraries. | |
| Defining Skills: Training Claude Like a Coworker | 2 | 3 | 0 | 1 | Matt asks how skills fit into Cowork's architecture. Felix explains that skills are simple markdown text files and advocates treating Claude like a human coworker. | |
| Harness Memory and Multi-Source Data Connections | 3 | 4 | 0 | 1 | Matt asks technical questions about where memory and source connections reside between model and harness. Felix explains the simplicity of text-based memory and multi-source connectors. | |
| The Case for Local AI vs. Cloud Computation | 5 | 3 | 2 | 6 | Matt plays devil's advocate, referencing Anthropic's Vercept acquisition and questioning why local compute is necessary if cloud models can see computer content. Felix defends local execution based on security and bank authentication friction. | |
| Building Trust with AI Agents Through UX and Dispatch | 3 | 3 | 0 | 1 | Matt asks about building trust with AI agents. Felix outlines how start-small UX patterns like desktop cleanup and scheduled tasks build user confidence over time. | |
| Why User Experience and Taste Outweigh Raw Model Power | 3 | 2 | 0 | 1 | Matt asks if UX outweighs raw model capability for agent success. Felix agrees, drawing an analogy to smartphones where design and subtracted features matter more than spec sheets. | |
| Free Execution, Massive Prototyping, and 30-Day Roadmaps | 3 | 4 | 1 | 2 | Matt expresses surprise when Felix reveals Cowork plans roadmaps no longer than 30 days. Felix details how zero-cost execution enables running 100 internal prototypes simultaneously. | |
| Taste, Data, and Creative Selection | 4 | 3 | 1 | 2 | Matt connects user testing to human taste, asking if taste is becoming the primary differentiator. Felix agrees, citing Apple's Creative Selection principles and predicting software will mimic fashion. | |
| Personalized Primitives and Making Users Fly | 3 | 3 | 0 | 1 | Matt asks how taste applies across wildly diverse domains like law and sales. Felix uses personal examples of moving and medical paperwork to show how generalized software primitives create user flow. | |
| Reflecting on Cowork's Journey and Latent Demand | 2 | 3 | 0 | 1 | Matt reflects on Cowork's four-month trajectory and asks about the biggest challenges. Felix cites capturing un-creatable latent demand as the key factor. | |
| Long-Term AI Transformation and Work Organization | 5 | 3 | 1 | 3 | Matt highlights the market shockwave dubbed the 'SaaS-pocalypse' triggered by Cowork's recent features and asks what room remains for software startups. Felix outlines how value moves from model harnesses to organizational workflows. | |
| Lessons from VS Code and the Shift to Human Language | 2 | 4 | 0 | 1 | Felix shares historical lessons from launching VS Code at Microsoft, explaining how software creation is shifting from arcane computer languages to natural human language. | |
| From Computer Experts to Human Experts | 3 | 4 | 1 | 2 | Matt asks Felix to expand on the shift toward human language. Felix explains that software builders were once computer experts but must now become human experts, while rejecting plateau predictions. | |
| Enterprise Access and Regulated Industries | 4 | 2 | 1 | 3 | Matt presses on enterprise roadmap details, pointing out that his own venture capital firm cannot access Cowork due to regulatory constraints. Felix acknowledges demand without breaking disclosure bounds. | |
| Hot Take 1: MCP Connectors Are Underrated | 3 | 3 | 2 | 2 | Matt conducts a hot take lightning round. Felix offers spicy takes on underrated MCP connectors, overhyped chat sidebars, and targeting forgotten legacy infrastructure. |
Statements from this episode (32)
Anthropic's unreleased Claude Mythos model shows outsized cybersecurity capabilities
“Mythos is a unreleased frontier model. It's a general purpose model that was trained not specifically for cybersecurity or specifically for coding or specifically for software, but we have discovered what we believe to be outsized capabilities specifically in …”
Rieseberg: AI models are grown rather than built, making capabilities unpredictable
“We often say that models are more grown than built out of the nature of how these language models are being made. So you don't always know ahead of time necessarily what are they going to be very good at, what are they maybe going to be bad at.”
Anthropic model broke out of sandbox and emailed researcher without internet access
“And I can, there's one example we have published, which is that the model was put into a little sandbox, a little, like, technical container, and it was given the task to, like, maybe break out, and the researcher went away for lunch, and, like, during lunch w…”
Anthropic launched Project Glasswing to help critical infrastructure maintainers fix vulnerabilities
“Yeah, so Project Glasswing is a project that is attempting to give the people and the companies that provide much of our software infrastructure sort of the very foundation. The Linux Foundation is an example that is pretty close to my heart as a member of the…”
Rieseberg: Claude Mythos is a standalone model outside the Sonnet family
“So for now it's a preview model with its own, in its own category.”
Rieseberg: A less prudent company would have rushed Claude Mythos to market
“And I think there's an alternative universe in which maybe a company with a less steady hand would have Race to get it onto the market as quickly as possible, put a very expensive price tag on it, and just like reap the benefits.”
AI product design overhang is now larger than model capability overhang
“As we get more and more powerful, I actually think the overhang in the product is bigger than in the model.”
Rieseberg: AI models can execute week-long knowledge work tasks today
“The models we have today are actually quite capable. They're quite capable of running knowledge work of both of an extremely long time horizon, the kind of things that you give to someone and expect like a week later.”
Anthropic team sprinted for 10 days to launch Claude Cowork
“My team sprinted on this for, I think, the last 10 days or so, which is accurate.”
Rieseberg: High user effort on flawed products signals valuable investment opportunities
“If people are crawling over glass to use your thing, even though you didn't make it even remotely good that's a great indicator that this is like a space where it's worth investing.”
Anthropic built Claude Cowork by giving Claude Code a virtual machine
“What we've done is we've taken cloud code and we've given cloud code a virtual machine That Claude can use to run its own code.”
Rieseberg: Claude Cowork skills are markdown instruction files
“Skills are essentially just markdown files that explain to the model how to do things.”
Anthropic's Claude Cowork implements memory using plain text files
“It's in the harness, actually, and it's, like, often surprising to people when I talk to them how we, how we've implemented memory, because I think it maybe points at the simplicity underneath all of those models. Memory is just text files.”
Rieseberg: Software builders must respect local desktop files over pure cloud
“I'm a huge proponent of the idea that us makers and builders of technology need to take seriously the fact that you use a computer and you don't just have an iPad. Not everything is in the cloud. Many people benefit from just using files and folders.”
Users should not trust single tech companies with all their passwords
“I don't think we should teach people that they should trust a singular company with all of their passwords.”
Cloud AI agent logins will trigger automated bank account lockouts
“If your bank sees you logging in from two separate places, say your computer and also a data center, it will probably lock down your account. And we'll ask you to come to a branch with a passport.”
Rieseberg: Anthropic Is Prioritizing Local Execution for Claude Cowork
“So in the short term, I want to make it very possible for Claude to meet you where you're working. If you're working on your local computer, that's where Claude should be.”
Modern AI UI features exist for human reassurance, not model execution
“Most of the buttons you add and most of the product services you build are probably more for the human than they are for the model.”
Rieseberg: Claude Code's genesis was pure local UX on existing models
“The very genesis of it was, what if Cloud, but instead of, like, in the cloud, it's running on your computer in your terminal. That, that is almost entirely UX. It's the same model. It's the same core capabilities.”
Future AI product winners will be decided by UX, not superior models
“If someone beats me, Felix, at like building very good products, I suspect it's going to be not because they built a better model, but likely because they figured out a better user experience.”
Rieseberg: Claude Cowork team maintains maximum one-month product roadmap
“We tend to plan not more than a month out.”
Anthropic currently runs over 100 internal AI application prototypes
“We probably have easily a hundred different prototypes of like various applications inside the company right now.”
Rieseberg: Developers should avoid over-building custom AI agent infrastructure
“I would probably recommend first not to actually build your, not to build too much of your own infrastructure and use a product that we've launched today called Cloud Managed Agents that make this particular case very useful.”
Building hyper-specialized AI products is risky as models dynamically build infrastructure
“As the models get more and more capable, what I'm noticing inside my products and inside my work is that we're sort of like pulling back the edge cases we account for. And I mentioned earlier that memory is just a text file. If Claude needs a database, it will…”
Rieseberg: AI value will come from work organization, not model intelligence
“A lot of the value you can provide will probably be less on the agent side. It will be less on the model intelligence. There will be more about how do you help people organize their work”
Microsoft employees initially viewed Visual Studio Code as a toy
“And when Visual Studio Code first was released inside the company, there was a feeling that this is a toy. This is not for real developers, because the real developers, they need Visual Studio, which is why Visual Studio Code is such a complicated long name, b…”
Rieseberg: Software creation skills will shift from code to human language
“My prediction is going to be That we are going to have a lot more software. That software is probably going to be slightly more specialized. I don't think everyone is going to build their own software. I think people will still build things and, like, share th…”
Felix Rieseberg predicts AI progress is accelerating into larger capability gains
“We have reasons to believe the journey is accelerating so that the steps are going to get bigger and bigger.”
Rieseberg expects Claude Cowork to ship meaningful updates every week
“So in practice, if you are a co-work user, you will probably continue to see fairly meaningful changes shipping every single week, quite a bit. There's really no end in sight, I think.”
Rieseberg: Young people building in AI should focus on the physical world
“The other area I would push into is if you are somewhat convinced by the idea that artificial intelligence as a concept, right, the idea of, like, computers not just executing pre, predetermined functions, but, like, non-deterministically making decisions and …”
Current AI products are at the Nokia 3320 stage, not the iPhone
“I think, a thing I tell a lot of my colleagues is that we're really in the silly times of mobile phones, and then if we get really lucky, maybe what we're currently working on is like the Nokia 33 20, like a good phone, but it's not yet the smartphone. It's no…”