Product
Visual recall, and the exception we argued about
One narrow case where text is not enough — and the rules we put around it before shipping a single frame.
August 17, 2026
Topic
ShogunAI product notes
How the product actually works: the execution layer, meeting memory, visual recall, and the rules around each of them.
What we publish and how we verify it
ShogunAI Editorial covers private AI memory, work context, local-first architecture, and the execution layer that turns an answer into finished work. We write for people evaluating how an assistant should remember, retrieve, and act across real tools.
Product claims are checked against the current implementation. Market claims link to named sources whenever a source exists, and investor opinions are labeled as opinions rather than presented as settled facts.
- Clear definitions before product claims
- Named sources for statistics and quotations
- Visible limits, privacy boundaries, and update dates
Product3 min read
Meeting notes without the recording
Meetings become part of your memory instead of a folder of audio files — and it is included in every plan, trial included.
August 17, 2026Read more →
Product3 min read
Bring your plan, not just your keys
Pro no longer requires an API key. ShogunAI can run its agent work inside the assistant subscription you already pay for.
August 17, 2026Read more →
Product1 min read
How to search across Slack, Gmail, and Notion
A practical workflow for finding the decisions and context that are scattered across Slack, Gmail, and Notion.
August 12, 2026Read more →
Product4 min read
The execution layer
Memory is table stakes. Acting on it — across the tools you already use — is the product.
July 18, 2026Read more →
Research behind the questions
These independent papers provide background, not evidence of ShogunAI’s performance or endorsement. For product behavior and limits, read the linked product guides.
- Lost in the Middle: How Language Models Use Long Contexts
- Nelson F. Liu et al. · 2023
- In the tasks studied, moving relevant information within a long input changed model performance. A larger context window alone did not ensure reliable retrieval.
- Generative Agents: Interactive Simulacra of Human Behavior
- Joon Sung Park et al. · 2023
- This simulation research combines stored experiences, reflection, and retrieval to plan agent behavior. It provides background on memory architectures, not a benchmark of workplace assistants.
