Distillate is a research tool for turning ML questions into autonomous agents that run experiments and feed the results back into later work. It centers on a workflow that links reading papers, generating experiment ideas, running agents, and extracting insights from the results.
The interface exposes commands for that loop: /conjure creates a research agent from a described question, with Nicolas setting up a git repo, drafting a prompt, and giving the agent a time budget. /steer lets a user adjust goals or change direction while an agent is running. /assay compares runs side by side to see what moved the needle, /distill extracts key findings from a full session history using local logs, and /survey gives an overview of new runs, breakthroughs, stuck agents, and other experiments that need attention. It also includes /transmute for turning paper insights into experiment ideas, /brew for syncing papers from Zotero and generating summaries, /forage for browsing trending research from HuggingFace and adding papers to a queue, and /tincture for extracting methods, results, and implications from a single paper.
Distillate is available on macOS and Windows, with installation commands shown for both the CLI and desktop app. The page says the CLI powers everything and the desktop app is a window into it. It also mentions connectors for Zotero, email, Obsidian, and reMarkable, along with support for reading and highlighting on reMarkable, iPad, and desktop. Extra workflow features include dashboards, charts, weekly email digests, PNG chart export, JSON state export, and automatic backups. A free account is available with no credit card required, and the core is open source. The page also states that it is powered by Claude Code and that it is a research preview.
In the Frameworks & runtimes space, Distillate takes a focused approach. It focuses on automating and orchestrating machine learning research experiments and extracting insights from papers. Distillate is an open-source project aimed at machine learning researchers. The project is open source (MIT). It runs on the web, the command line, macOS, and Windows.
Distillate first shipped in 2026. The project is developed in the open on GitHub with 72 stars and 2 commits in the last 90 days. Among its 5 catalogued features are auto-experiment orchestration, research agent spawning, and paper insights extraction. Distillate is currently in beta.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Closest matches by what these projects do