This is post zero. A short note about what this thing is, and what it isn’t, before the first real post lands.
What it is
An engineering blog. First-person, technical, fundamented. Decisions explained — not announced. Trade-offs surfaced. The kind of writing the people at Linear, Vercel, and Stripe publish on their own engineering blogs. The kind that Geoffrey Litt and Simon Willison write in long form on their own sites.
It comes in three shapes:
- Engineering posts — long-form deep dives. Architecture decisions, with the rationale and the alternatives we rejected. Monthly to bi-weekly. “Why we chose hierarchical paths over pure embeddings.” “How RiLiGar Memory handles temporal decay.”
- Notes / changelog — what shipped this week. Short. A bullet list, sometimes a paragraph. “New integration path: Crew AI.” “Bridge type added:
derives_from.” - Field notes — observations on the ecosystem when something interesting moves. Irregular. “On the SEP-1865 specification.” “Re-reading Letta’s paper: what holds up, what doesn’t.”
What it isn’t
It’s not “we hit $2.3k MRR this month, here’s what I learned.” It’s not “failed launch on Reddit, lessons in vulnerability.” It’s not “10 lessons from building my SaaS.” The number-spectacle and the lesson-listicle don’t fit the kind of attention we’re trying to earn.
We’re not writing for peers. We’re writing for the engineer who needs to understand a decision because they’re about to make a similar one, or because they’re using something we built and want to know how it actually works.
The cadence
One to two posts per week. Sustainable. A typical month: four notes, one or two engineering posts, zero to two field notes.
The rule: never publish without a thesis. No filler, no recap-as-content. If we don’t have something to defend, we don’t post. Volume without substance corrodes the brand faster than silence does.
What’s signed by a human
Anything with an opinion. The thesis, the personal voice, the contraintuitive line, the confession of an error — those don’t get drafted by a model and rubber-stamped. The model assists with research, structure, tone-checking against the canonical brand documents, SEO, distribution variants. It doesn’t carry the opinion. The opinion is always human.
If a post ever shows up here that reads like commodity AI prose, that’s a bug. Tell us.
Subscribe via RSS
We’re keeping it simple — there’s an RSS feed and that’s it for now. Newsletter and comments come later, if they come. The feed is at blog.riligar.com/rss.xml.
The first real engineering post is queued. It explains why our memory layer treats hierarchical paths as primary structure and embeddings as the index — not the other way around.
— RiLiGar