Nov 29, 2017 · 57m · y-combinator
Microbes, Robots, and Ambition - Robin Sloan on His Novel Sourdough · Y Combinator
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Author Robin Sloan joins hosts Craig Cannon and Kat Manalac to discuss his novel Sourdough, exploring machine learning applications in writing, digital ephemerality, kitchen automation, and the evolving culture of Silicon Valley.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. How this is scored →
speaking balance: gold is the partners, purple is the guest (3 minute bins)
Robin Sloan directly attacks a conventional piece of creative writing advice, telling the audience that anyone saying a story seeks its own vessel is 'full of shit' because form precedes content.
Hardest push from the partners ▶ 30:50 Craig challenges the premise of leaving no permanent digital traceCraig Cannon pushes back against Robin Sloan's argument that preserving conversations is weird and antisocial by noting that he intentionally records, transcribes, and indexes his entire life.
Biggest teaching moment ▶ 43:20 Sloan demonstrates how neural networks should be integrated into writing toolsRobin Sloan educates the hosts on moving past simple command-line neural network generation, explaining how he designed a live interactive tab-completion workflow in Atom to keep the human in the loop.
The partners hold their own ▶ 40:30 Craig demonstrates technical grasp of Magenta and style transfer pitfallsCraig Cannon displays deep familiarity with creative AI tooling by referencing Google Magenta and explaining why simplistic neural style transfer alienated visual artists.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The partners as informed peer | Guest teaching | Guest disagreement | The partners pushing back | Why |
|---|---|---|---|---|---|---|
| Machine Learning and the Fictional Music of Sourdough | 3 | 4 | 0 | 0 | Craig Cannon sets up the conversation by asking about Robin Sloan's machine learning audio experiments. Sloan provides a detailed technical breakdown of training SampleRNN in Torch on Croatian Klapa music to synthesize fictional singing for the Sourdough audiobook. | |
| Researching Robotic Arms & Technical Accuracy in Fiction | 3 | 2 | 1 | 1 | Craig brings up Spike Jonze rejecting technical advisors on film sets, contrasting it with technical accuracy in fiction. Sloan explains his research into robotic arms and why organic kitchen manipulation remains a difficult robotics problem. | |
| Inspiration for Sourdough and Bridging Tech with Traditional Crafts | 2 | 3 | 2 | 0 | Kat Manalac asks why Sloan focuses on the intersection of tech startups and traditional crafts like breadmaking. Sloan reframes the conflict, passionately rejecting the narrative that new technology must replace ancient crafts. | |
| Genre Classification: Sci-Fi Literary Fiction and Optimistic Futures | 3 | 2 | 1 | 0 | Craig and Kat discuss the modern prevalence of dystopian narratives like Black Mirror versus Sloan's optimistic sci-fi literary fiction. Sloan argues that writing characters people want to spend time with offers a necessary alternative to bleak cynicism. | |
| Children's Literature and Microbiome Research Influences | 2 | 4 | 3 | 0 | When Craig asks if Sloan ever considered building an app alongside his book, Sloan bluntly rejects the common industry dogma of finding the 'right vessel' for a preexisting story. He then illustrates the fragile impermanence of apps compared to physical books through his lost iOS project Phish. | |
| Preserving Digital Culture and Auto-Deleting Tweets | 3 | 2 | 1 | 2 | The conversation shifts to digital archiving, code preservation, and Sloan's practice of auto-deleting his tweets. Sloan likens Twitter to an ephemeral cafe conversation rather than a permanent congressional record, while Craig humorously notes he indexes his entire life. | |
| The Golden Age and Decline of Blogging | 2 | 3 | 0 | 0 | Kat asks about Sloan's early blogging work on Snarkmarket and whether social platforms killed it. Sloan reminisces about the mid-2000s blog ecosystem, RSS curation in Google Reader, and the lost craft of daily public thought. | |
| Reflections on Twitter: 280 Characters and Interface Design | 3 | 2 | 1 | 1 | Craig asks about the controversial expansion of Twitter to 280 characters. Sloan recounts an early college lesson about UI changes on the New York Times site and defends 280 characters for allowing users to construct dialectical thoughts. | |
| Machine Learning as an Interactive Tool for Creative Writing | 4 | 3 | 1 | 1 | Craig and Sloan discuss generative ML tools, with Craig citing Google Magenta and the limits of style transfer. Sloan outlines his philosophy of building interactive, human-in-the-loop ML tools by detailing his custom Atom text editor autocomplete plugin. | |
| The 'Child of Hogwarts' Quote and Books as Social Objects | 2 | 2 | 0 | 0 | Craig brings up Sloan's popular 'child of Hogwarts wanting to be sorted' line from Sourdough. Sloan discusses how isolated book quotes function as viral social media objects while maintaining that deep reading remains insulated from fast-paced online behavior. | |
| Epic 2014: Retrospective on an Early Viral Flash Video | 2 | 3 | 0 | 0 | Craig asks about the origins of Sloan's 2004 viral Flash video Epic 2014. Sloan describes building the media-future mockumentary with Matt Thompson at Poynter and the grassroots global web mirroring required to distribute video before modern platforms. | |
| Shifting Tech Narratives from Penumbra to Sourdough | 3 | 2 | 1 | 0 | Kat asks how changing tech journalism tone influenced the writing between Mr. Penumbra and Sourdough. Sloan explains toning down pure techno-optimism to reflect the darker realities of startup labor while intentionally steering clear of lazy nihilism. |