Felix Rieseberg

The Oxford Internet Institute lists Rieseberg as a former MSc student from October 2010 to July 2011, with a completed thesis on how temporal sorting affects perceived information relevance. His later engineering career moved through Microsoft, Slack, Stripe and Notion before Anthropic. His personal timeline dates these roles to 2011–2016, 2016–2021, 2021–2023 and 2023–2025 respectively.[1][2]

Making desktop software easier to build and explore

Rieseberg has helped maintain Electron since 2016 and created Electron Fiddle, a tool for trying small examples with different Electron versions. His windows95 and macintosh.js applications package older operating-system experiences inside desktop apps, using existing emulation engines. At Slack he worked on the cross-platform desktop client before managing the team; at Notion he worked on its desktop applications and underlying web infrastructure.[1]

A small language model as a learning instrument

Language Model Builder combines an interactive introduction with a Mac training workbench. A learner can choose tokenization and data, watch pretraining loss, fine-tune on examples, apply preference training and chat with the resulting model. The app exposes intermediate behavior such as token probabilities and saved checkpoints. Its goal is to make model construction visible to people without coding or machine-learning experience, using a deliberately small model that can run locally.[3]

In his ambition tutorial, Rieseberg distinguishes doing familiar work faster from raising the level of the goal, investigating a task more deeply or borrowing methods from another field. His example of moving from fixing individual crashes to maintaining a crash-free application depends on providing crash reports, source mappings and test machines. The lesson is that a broader request works better when the surrounding workflow gives the model the information and verification tools it needs.[4]