Ignore previous directions number 2
Welcome to the second edition of the newsletter, also written mostly on a train, to London this time.
Nature
The oxeye daisies are out everywhere in their hordes.
Events
Last week I went to Open Source Founders Summit in Paris. This was the second event for this conference, with some familiar people from last time and many new ones. Thanks to Emily Omier and Remy Bartot for all the work organising the event. As the event was under Chatham House rules, I am not going to give a lot of detail, but just cover some impressions.
I spent a lot of time in the sales and marketing breakouts with Lisa Cao, Matt Barker, William Morgan and others. I am going to write a lot more about open source business soon, but some thoughts here. Open source can give you advantages in sales and marketing, with inbound being potentially strong, and community marketing and adoption being great drivers. But you still have to sell, and there were people who wanted to avoid doing that, and just run technical teams. I think we may have dissuaded them by the end fo day two. In many ways this shows the reasons for having a business focused open source conference, as people often start open source businesses more or less by accident when a project becomes successful, rather than as a strategic design, and the business side is often new to them. The strong message from people who had done this more than once is to be more strategic about open source and build a strategy up front.
A number of people had set up non profit or similar organisations, or otherwise restricted themselves. Often there were good reasons for these, and people want to build things in ways that are aligned with how they want to live, but running a business on open source is hard, and limited how you can operate in future reduces your options and may make it even harder, especially as the environment changes or you want to make organisational changes to adapt to circumstances.
Open source can increase competition, especially competition that does not have the costs of supporting the project so can undercut you. This is where a lot of the goodwill tends to break down in communities about projects. Building a business on open source you need to give yourself some advantages to keep the business sustainable too.
People
I caught up with Rod Johnson at OSFS and talked about his new project Embabel. The difference between this and other agent frameworks is that it is based on Goal Orietned Action Planning (GOAP), which is a technique largely used for building game AI. Essentially you take a set of (typed) actions that the AI can perform, each with costs attached, and make a plan which you can later refine. In gameplay you might have an AI character who has a goal that needs a sword and various actions available that can lead to a sword, like accumulating enough coins to buy one. You can imagine a similar setup with the actions being typed MCP servers or similar tools, and planning to drive an outcome. Most agentic frameworks leave the planning part as an implicit function the model needs to drive without explicit structure, so this is an interesting different approach.
Reading
- I am disappointed by the AI discourse by Steve Klabnik
- Stop building AI tools backwards by Hazel Weakly
Take control and build the stuff you want might be the summary of this weeks reading. I need to start on my unread book pile next week...