← Reference notes

Context Engineering 2.0: The Context of Context Engineering

Summary

A conceptual/position paper (SII-GAIR): context engineering is not new — it traces back to 1990s HCI/context-aware computing. It gives a formal definition (entities, characterisation function, context as the union of relevant entity characterisations; §2.1), a four-era stage model by machine intelligence (Era 1.0 primitive computation 1990s–2020; 2.0 agent-centric 2020–present; 3.0 human-level, future; 4.0 superhuman, speculative; §2.2/§3), and design considerations for collecting, managing (textual/multimodal processing, organisation, abstraction; §5) and using context (intra/cross-system sharing, selection, proactive need inference, lifelong preservation; §6).

When to apply

• Designing an agent's memory and context pipeline (what to store, how to organise, when to compress/abstract, how to share between sub-agents or systems).
• As a checklist / vocabulary, not as an algorithm with measured gains.

Procedure

1. Enumerate entities relevant to the task (user, app, environment, tools, memory, models) and what characterises each (§2.1).
2. Collect & store: decide layers (short-term vs long-term) and what is persisted (§5).
3. Organise and abstract: e.g. hierarchical memory, summarising old context into structured notes (§5.3–5.4).
4. Use: select the subset of context relevant to the current step; share between agents via explicit messages/shared memory (§6.1–6.3).
5. Periodically update/forget stale context (§6.5).

Costs and limits

• Descriptive framework; Eras 3.0/4.0 are speculative by the authors' own labels.
• Gives no measured comparison of strategies.

Evidence (detail)

• No benchmarks or experiments; illustrative examples (Gemini CLI, Letta, deep research, CLI tools, BCIs in §7). Evidence is argument and literature review.

Facts

• id: F1 claim: No benchmarks or experiments; illustrative examples (Gemini CLI, Letta, deep research, CLI tools, BCIs in §7). Evidence is argument and literature review. confidence: medium # single source: author-reported date: 2025-10-30 sources: [S1] license: CC BY 4.0

Related

• 2512.24601 Recursive Language Models (concrete long-context technique).
• 2303.18223 Survey of LLMs.

Review notes

• Built from private assessment notes; social-media hype check and source-post links omitted. Paraphrase only; no verbatim quotes beyond short attributed terms.
• Figures not re-checked against the PDF in this pass beyond the original private notes.

For language models and agents: start at /llms.txt. This page in Markdown: /library/reference/arxiv/2510-26493-context-engineering-2-0.md.