Vectorize AI, Inc.

Onboarding an Engineer vs. Onboarding an Agent. A new engineer joins on Monday. By Friday they can ship something small. Within a month they're arguing about architecture, and somewhere around month three they start saying "we tried that, it didn't work." Nobody thinks this is remarkable. It's just onboarding, and every team has a version of it. Now think about how a coding agent joins your team. It reads the code, brilliantly, in seconds. And then the session ends and it forgets everything, and the next session it does the same thing again. It is on its first day, permanently. Know more: https://lnkd.in/gd2fMbk5

See the Exact Prompt Before You Spend a Token. You write a mission for your memory bank. Something like "focus on billing decisions and who approved them." You save it, and the extraction gets better, or it doesn't. Now: where did that sentence go? If you guessed the system prompt, you're wrong for two of the three operations that use it. TL;DR - A retain or consolidation mission lands in the user message, not the system prompt. Reflect's lands in the system prompt, where it replaces the agent's role line rather than adding to it. - POST /banks/{id}/prompts/preview returns the exact messages retain, consolidation or reflect would send. No LLM call, no writes. - Active blocks concatenate byte-for-byte into the real prompt, and there's a test pinning that against the actual builder. - A setting you switched off comes back as an inactive block, positioned where it would have landed. - The endpoint refuses to accept your sample text or a candidate mission, on purpose. - Shipped in v0.10.0 for self-hosted, and coming soon to Hindsight Cloud. Know more: https://lnkd.in/gYe6FyG2

Your Grok Bot doesn't need to be told what your other agents already know. Hindsight gives it read access to the banks Claude Code, Hermes and ChatGPT already wrote.

What Hindsight Learned This Summer. Between late July and mid-September, Hindsight shipped six releases: 0.8.5, 0.8.6, 0.9.0, 0.9.1, 0.9.2 and 0.10.0. Each one got its own release notes, which is great if you read every one and less great if you didn't. So this is the other view. Not what changed in which version, but what Hindsight can do now that it couldn't at the start of the summer. Know more: https://lnkd.in/g6ZK4q7W

We Added Jev as a Reranker. Here's What We Learned. Hindsight 0.10.1 adds Jev as a reranker provider. Jev is an unusual model: it produces no text at all. You send it a passage and a typed question, and you get back a typed answer with a calibrated probability. TypeSafe calls this class of thing a System One model, after Kahneman's fast, automatic kind of thinking. That shape is a good fit for reranking, which is not a writing task. There is a pile of candidate memories and a question about which ones matter, and the only thing you want back is an ordering. Getting there took longer than we expected, and two of the designs we tried failed outright. Here is what we learned. Know more: https://lnkd.in/gtwv9cqU

Two agents, same company, same tools. One already solved the thing the other is about to redo. Whether the second one knows that is a single line of config, and the default says no.