I am Brian Galvan, and I built this. Distillation Technologies is not a real company, which you have probably worked out by now. It is a satire of the AI industry, and specifically of the argument that has been running all year about which laboratories have been copying which other laboratories.
The joke is the surface. Everything underneath it is real, and building it was the actual point.
Why I built it
I use projects like this to practise. Reading about a technique and shipping something with it are separated by a surprising amount of detail, and the only way through that detail is to build the thing and watch where it breaks. A site with a real domain, real deployment, and real search exposure breaks in more interesting ways than a tutorial does.
This one let me work on several things at once.
Autonomous agents
This is the part I care most about. The newsroom on this site is written by an agent I built. Once a week it scans global coverage of model distillation, decides whether anything genuinely new has happened, researches it, writes the article, verifies every source it cites by actually fetching them, commits to GitHub, rebuilds the site, publishes, and reports back to me on what it did.
There is no approval step. It is a fully autonomous reporter, and the hardest behaviour to get right turned out to be the ability to decide that nothing happened this week and publish nothing at all. An agent obliged to produce something every week will start inflating trivial commentary into events.
More broadly I work on LLM and agent systems, including graph databases as memory and retrieval layers for models, which is where a lot of my attention sits at the moment.
I have written up the architecture in full, including the three bugs that only turned up once it was actually running: ALEMBIC, the autonomous reporter.
The marketing side
I have spent my career in martech and DevOps, and search is the part of marketing that never sits still. It went from SEO to SEM, and it is now moving into GEO and AEO as people stop clicking through to pages and start accepting answers from AI systems directly.
That is a change in human behaviour, not a channel adjustment, and I wanted somewhere to practise the TTPs against it. Structured data, an llms.txt that tells crawlers what is fact and what is fiction, content written to survive being quoted with no surrounding context, the technical build, the creative side, all of it. This site is that sandbox.
What I do the rest of the time
I work with several companies. I manage martech and AI initiatives for a law firm. I also work with a private aviation company, where we are building the future of aviation maintenance platforms and bringing technology into an industry that has resisted it for a very long time.
Alongside that I run a small boutique application and app development company. It is referral only, and a good share of my time in it goes on learning and on keeping everything running.
Where my head actually is
Underneath all of it, what I find most interesting is the technology powering these models rather than the products built on top of them.
I have come around to Yann LeCun's view that autoregressive language models are not the whole path, and that world models are likely to be what carries us to the next level. I spend a lot of time reading and thinking about what becomes possible after LLMs, and what the generational jump past them actually looks like.
My honest position is that we have not fully conceived of it yet.
Get in touch
If you want to talk about any of this, the philosophy of AI, how civilisation is being reshaped by it, or world models and what they might mean for AGI, I am on LinkedIn and happy to hear from you.
And if you came here for the satire, the Learning Center is the part of the site that is entirely serious, and the newsroom is where the agent files its copy.