Dosu

📕LLM Wiki Webinar with Brace Sproul Devin Stein Jeff Huber is now on YouTube! Two of my favorite insights from this webinar: "Wiki as a cache" - Devin Stein: basically, the purpose of a wiki is to keep things that are commonly looked up or accessed more top of mind/readily available. What goes in the wiki should be updated and organized with that in mind "Wiki is a set of hyperlinked pages" - Jeff Huber. The evolution of memory (in my view) has gone from: 1. single string 2. file 3. set of files in a directory (wiki) as you think about scaling up - as files grow, the file structure matters less, and you need links between pages. The internet is a set of hyperlinked pages - so that definitely scales! Watch it here: https://lnkd.in/ggf5xE7S

models come and go, knowledge (can) last forever. join us on Thursday to learn how to create, maintain, and leverage knowledge for agents: https://lnkd.in/ejkQswUW

The first 6-months of 2026 were are all about Token Maxing. The next 6 months are all about Token Efficiency. One of the best ways to improve token efficiency is via "LLM Wikis" or compiling agent learnings into re-usable knowledge artifacts. LangChain recently launched OpenWiki to help teams generate their own LLM wikis. Dosu provides a shared Knowledge Cache for agents across teams at organizations. Now we're teaming up for an emergency webinar this Thursday to discuss what "LLM Wiki" really means and if it's actually the right abstraction for agents. If you're wiki-curious and want to learn how your team can improve token efficiency by up to 50%, come join Harrison Chase Brace Sproul and me for a heated discussion this Thursday. https://lnkd.in/ejkQswUW

Does your agent need a wiki? Should humans and agent use the same wiki? Should you have one wiki or many wikis? If you're wiki-curious, join Brace Sproul Harrison Chase and I on Thursday! Dosu 🤝 LangChain https://lnkd.in/ejkQswUW

Went looking for best practices on building GTM & Sales Ops agents to automate the repetitive work that eats the days of Devin's, Taylor's, George's, Jono's and mine. But I walked away with clear outside validation on why Dosu exists.. 😅 After getting hooked by what Vercel has built, I went deep on Tom Occhino's (their CPO) breakdown of how they built their internal deal agent. I was there for the playbook: pick low-value repetitive work, document the process in detail, hand it to a coding agent, keep a human in the loop. Then he said something that had nothing to do with sales agents and stopped me cold: "Documentation is the bottleneck, not the model." His point: the hard part of good agent output was never the AI. It was being able to write down, precisely, how the work actually gets done. Teams that never documented their processes are about to find out that is the real blocker, no matter which model they pick. I sat there realizing he had just described, from the outside, the exact problem Dosu was built to solve. Not for sales agents. For coding agents working on a codebase. And it goes one step further than even he said. Writing it down once is not the finish line. The moment you document how your codebase works, the code moves. Now your agent is reading a confident, detailed, slightly wrong description of a system that changed last week. Stale context is worse than none, because the agent trusts it. So there are two bottlenecks. Documentation existing at all. And documentation staying true as the code changes underneath it. The second one, kept fresh and trustworthy on every run, is the reason Dosu exists. LOVE when I go looking for something else and run into the kind of validation you can't manufacture. 🫶

It's Dosu Drop Day! ☔ This month, we rebuilt how you configure Dosu, and the inspiration came from a Borges short story written in 1941. In "The Library of Babel," an archive holds every possible 410-page book. Infinite knowledge, infinite misinformation, and mostly noise. Borges' point is that organization is the prerequisite for useful knowledge. So we reorganized Dosu with this principle in mind. 📚 Spaces are now Libraries. A Library controls what knowledge Dosu can access and maintain, and each one holds Sources — your connected integrations. Set a Source to Read for context, or Monitor to watch your code and keep that knowledge fresh as it changes. Your existing Spaces migrated automatically. 🦾 Agents. A new section that lets Dosu chat with you in Slack, MS Teams, or GitHub Issues and Discussions. @Dosu to "save this thread," "generate a changelog," or "who shipped this feature?" Each Agent draws its context from a Library you choose. Other things worth sharing...we wrote up how he uses Templates to draft our weekly Discord posts, and simpler usage-based pricing is on the way (usage, not seats). You can check out the full Drop here: https://lnkd.in/gsDP7gef #ai #documentation #knowledgeinfrastructure #agents #llm #developers #MCP

The new website from dosu is one of my favorites in a long time: https://dosu.dev/