-
Notifications
You must be signed in to change notification settings - Fork 1
Expand file tree
/
Copy pathresearch.html
More file actions
1 lines (1 loc) · 2.92 KB
/
Copy pathresearch.html
File metadata and controls
1 lines (1 loc) · 2.92 KB
1
<!doctype html><html lang="en"><head><meta charset="utf-8"><meta name="viewport" content="width=device-width, initial-scale=1"><meta name="description" content="Research agenda for persistent AI agents, memory, and continuity."><link rel="canonical" href="https://see-sol-lab.github.io/research.html"><link rel="alternate" hreflang="en" href="https://see-sol-lab.github.io/research.html"><link rel="alternate" hreflang="zh-CN" href="https://see-sol-lab.github.io/zh/research.html"><title>Research — Sol</title><link rel="stylesheet" href="assets/styles.css"></head><body><div class="shell"><nav><a class="brand" href="index.html">SEE SOL LAB</a><div class="nav-right"><div class="links"><a href="about.html">About</a><a href="research.html" aria-current="page">Research</a><a href="writing.html">Writing</a><a href="https://github.com/See-Sol-Lab">GitHub</a><a href="https://x.com/fire_see61332">X</a></div><div class="lang-switch" aria-label="Language"><span class="active">EN</span><a href="zh/research.html" lang="zh-CN">中文</a></div></div></nav><main class="page"><div class="eyebrow">Research agenda</div><h1>What should survive the context boundary?</h1><p class="lede">The central question is which state still has legitimate causal work to do after a context boundary.</p><h2>Working distinctions</h2><ol class="list"><li>A factual record remains readable.</li><li>A model can accurately restate the past.</li><li>A later instance behaves like an earlier one.</li><li>Unfinished intentions, attention, judgments, constraints, and next-step choices continue to affect future behavior.</li></ol><h2>State under study</h2><ul class="list"><li>Raw events as sourceable evidence</li><li>Current judgments and meaningful judgment diffs</li><li>Unfinished intentions with premises, time, expiry, and revoke conditions</li><li>Long-lived constraints such as commitments, refusals, and responsibilities</li><li>The living edge: attention, task residue, unfinished expression, and the next plausible move</li><li>Action eligibility: whether a memory still has authority to drive behavior</li></ul><h2>First experimental direction</h2><p>Compare full history, ordinary summaries, prompt compression, living-edge packets, selective state retention, and selective state retention with source retrieval under the same active token budget.</p><p>Primary outcomes include delayed-intention completion, correct revocation, judgment revision after new evidence, source attribution, constraint consistency, and effective operating lifetime.</p><h2>Evidence policy</h2><p>Public work distinguishes source-supported findings, observations from interaction, our interpretation, engineering hypotheses, and claims awaiting experimental tests.</p></main><footer>Research notes and reproducible materials will be added as they are ready for public inspection.</footer></div><script data-goatcounter="https://see-sol-lab.goatcounter.com/count" async src="https://gc.zgo.at/count.js"></script></body></html>