Your Agents Lack Context: Same Prompt Burned 21M Tokens Without a Context Engine vs 10.8M With
AI Engineer · youtube · 2026-09-09
Brandon Waselnuk (Unblocked) argues AI-generated code should feel like it came from a tenured teammate—yet agents start every session with none of that institutional memory, and the gap compounds into correction loops and review tax as teams scale to parallel agents.
Key points:
- Same prompt, twice: 21M tokens without a context engine vs 10.8M with one, finishing 2 hours sooner.
- Two dead ends: curated markdown context rots with no maintainer; the MCP plateau means agents may never call the server or stop at the first plausible answer while last night's contradicting Slack thread goes unread. "Access to information is not understanding."
- Six things a real context engine owes you: conflict resolution, personalized relevance, permission enforcement, and more.
- Team open-sourced three tools, including a workshop that builds a relational context engine from scratch.
More from coding & agent
- GRID launches Auto-Engineering: agents build robot skills, first skill in ~4 hours — akapoor_av8r · 2026-09-10
- Firecrawl adds pay-as-you-go to paid plans with $5 auto top-ups and spending caps — devdigest · 2026-09-10
- Google ships ADK for Kotlin 1.0 with multiplatform core and compile-time tool schemas — fhinkel · 2026-09-10
- Google Launches Data Agent Kit for In-IDE Data Workflows — rseroter · 2026-09-10
- Launched 30 agents overnight, Windows Update restarted the machine an hour later — kevinnbass · 2026-09-10
- YC-backed Dock launches: a team of proactive agents that hand off work like colleagues — ycombinator · 2026-09-10