A retrospective on a thread of posts about personal AI ownership, what's clarified, what's still murky, and what I'd revise about the earlier framing now that more of the substrate exists.
Three weeks after the BSL announcement, the OpenTF Manifesto became OpenTofu, now under the Linux Foundation. The fork happened faster than any IaC story I can remember. Here's what actually shipped, what migration looks like, and why this is structurally healthy for IaC.
HashiCorp moved Terraform from MPL to BSL on August 10. The license itself is narrower than the panic, but the panic is rational. Here's what actually changed, what it means for teams running multi-cloud Terraform in production, and what to do this week.
Meta released Llama 2 last week with a permissive commercial license. Five months after the leak made local inference possible, the legal layer just caught up to the technical one.
Anthropic shipped Claude 2 yesterday with a 100K-token context window. The capability is genuinely interesting; the architectural shift it forces is the bigger story.
ChatGPT plugins shipped a month ago. Function calling shipped last week. The substrate for agents that take action on your behalf just landed, and the second-order problems are going to be ugly.
If the goal is a working representation of how a specific person reasons in their domain, what's the actual stack? Walking through corpus, model, retrieval, eval, and where each piece breaks today.
The model providers scraped the open internet and called it fair game. The next phase needs an actual market for the data, and the structural pieces of that market don't exist yet.
If you could capture not just what someone knows but how they reason, and run that on local hardware, what would the artifact actually look like?
Capturing how someone writes is not the same thing as capturing what they know. Both are interesting capabilities. The conflation between them is going to cause problems.
GPT-4 dropped a week ago. The capability jump is real. The ownership question it surfaces, about the corpus you use to teach it, is sharper than the marketing suggests.
Meta released LLaMA to researchers on February 24th. Three days later the weights were on 4chan. The rest of 2023 will be shaped by what happens next.
