marketing-shibata50/claude-code-ultimate-guide
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

docs: add Spring Break 2x usage promo note + ICM evaluation

Browse source 
- Guide: documented Anthropic March 13-27 2026 promotion (2x limits
  outside 5-11am PT / all weekends), with CET timezone conversion for
  European users
- Resource evaluation: ICM Infinite Context Memory MCP server (score 3/5)
- CHANGELOG: both entries under [Unreleased]

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
Florian BRUNIAUX 2026-03-15 18:34:55 +01:00
parent 30e3916eb3
commit d9cff74d71
3 changed files with 212 additions and 9 deletions
Download patch file Download diff file Expand all files Collapse all files

1
CHANGELOG.md
View file

@ -8,6 +8,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/).
### Added
- **Spring Break promotion note** (guide line ~2395): Documented Anthropic's March 13-27, 2026 promotion — 2x usage limits outside 5-11am PT (peak hours) and all weekends, bonus usage doesn't count against weekly limits, applies to Free/Pro/Max/Team. Includes CET timezone conversion for European users (2x from midnight-13h and 19h-24h France time). Source: Anthropic support article.
- **ICM (Infinite Context Memory)**: New MCP memory server section after Kairn (~line 11365) — Rust single binary, zero deps, Homebrew install, dual architecture (episodic decay Memories + permanent knowledge graph Memoirs), 9 typed relation types, auto-extraction 3 layers, 14 editor clients. Score 3/5 — recommended as Rust-native alternative when Python dependency management is a friction point. Includes explicit license callout (Source-Available, free ≤20 people) and vendor-reported benchmark flags.
- **Comparison matrix update**: Added ICM column to MCP memory stack matrix (Runtime + License rows added for all tools)

113
docs/resource-evaluations/icm-infinite-context-memory.md Normal file
View file

@ -0,0 +1,113 @@
# Resource Evaluation: ICM (Infinite Context Memory)
**Date**: 2026-03-14
**URL**: https://github.com/rtk-ai/icm
**Type**: GitHub repository / MCP server
**Score**: 3/5 — Integrated
**Decision**: Added as section after Kairn in `guide/ultimate-guide.md` (~line 11365)
---
## Summary
ICM is a persistent memory MCP server from the rtk-ai team (same authors as RTK/Rust Token Killer). It provides a dual memory architecture: "Memories" (episodic, configurable decay) and "Memoirs" (permanent knowledge graph with 9 typed relation types). Distributed as a single Rust binary with zero external dependencies, installable via Homebrew.
### Key Points
- Single Rust binary, SQLite, zero deps — Homebrew install
- Dual architecture: episodic decay + permanent knowledge graph in one tool
- Hybrid search: BM25 30% + vector similarity 70%
- Auto-deduplication (>85% similarity blocked)
- Auto-extraction: pattern hooks, pre-compaction, session-start
- Supports 14 editors/clients (Claude Code, Cursor, VS Code, Windsurf, Zed, Amp, Cline, etc.)
- 52 stars, 55 commits as of March 2026
- License: Source-Available (free for individuals and teams ≤20; enterprise license required above)
---
## Scoring
| Criterion | Score |
|-----------|-------|
| Relevance to Claude Code users | 4/5 |
| Differentiation from existing content | 3/5 |
| Maturity / adoption signal | 2/5 |
| License openness | 2/5 |
| **Overall** | **3/5** |
---
## Comparison vs Existing Guide Content
| Feature | doobidoo | Kairn | ICM |
|---------|----------|-------|-----|
| Language | Python | Python | Rust (single binary) |
| Install | pip | pip | Homebrew |
| Episodic decay | No | Yes (biological) | Yes (configurable) |
| Permanent knowledge graph | No | Yes | Yes (Memoirs) |
| Auto-extraction | No | No | Yes |
| License | MIT | MIT | Source-Available |
Main differentiator: zero-dependency Rust binary lowers install friction for users who struggle with Python environments. Conceptual overlap with Kairn (knowledge graph + decay) is real but the runtime difference is meaningful.
---
## Benchmarks
All benchmarks below are **vendor-reported by rtk-ai** — not independently verified.
**Storage performance (1000 ops, 384d embeddings)**:
- Store (no embeddings): 34.2 µs/op
- Store (with embeddings): 51.6 µs/op
- FTS5 full-text search: 46.6 µs/op
- Vector search (KNN): 590.0 µs/op
- Hybrid search: 951.1 µs/op
**Agent efficiency (Haiku model, multi-session)**:
- Session 2: 29% fewer turns, 32% less input context, 17% cost reduction
- Session 3: 40% fewer turns, 44% less context, 22% cost reduction
**Knowledge retention (10 questions)**:
- Without ICM: 5%
- With ICM: 68%
Note: The knowledge retention benchmark uses a sample of 10 questions on Haiku — too narrow for generalization.
---
## Fact-Check
| Claim | Status | Source |
|-------|--------|--------|
| Storage: 34.2 µs/op | ✅ | README benchmarks section |
| Hybrid search: ~951 µs/op | ✅ | README benchmarks section |
| 29-40% turn reduction | ✅ present / ⚠️ vendor-reported | README — rtk-ai self-evaluation |
| 68% vs 5% knowledge retention | ✅ present / ⚠️ vendor-reported, n=10 | README |
| Source-Available license, free ≤20 | ✅ | LICENSE file |
| 9 Memoir relation types | ✅ | README full list |
| 14 supported clients | ✅ | `icm init` documentation |
| 52 stars | ✅ | GitHub as of 2026-03-14 |
No hallucinations detected. All figures present in the source README.
---
## License Note
Source-Available license. Free for individuals and teams of up to 20 people. **Enterprise license required for organizations above 20 people.** Contact: license@rtk.ai
This was flagged in the guide entry with an explicit callout. Teams should verify their org size before deploying.
---
## Integration Location
- New section: `guide/ultimate-guide.md` after Kairn (~line 11365), before "MCP Memory Stack: Complementarity Patterns"
- Comparison matrix updated: ICM column added with Runtime and License rows
## Upgrade Trigger
Revisit for 4/5 if:
- Benchmarks independently verified by community
- Adoption exceeds 500+ stars with sustained commit activity
- License changes to MIT/Apache

107
guide/ultimate-guide.md
View file

@ -2394,6 +2394,10 @@ Monthly cost estimate: $50-$200 for 5-10 developers
> **Note**: Anthropic's plans evolve frequently. Always verify current pricing and limits at [claude.com/pricing](https://claude.com/pricing).
> **Spring Break Promotion (March 13-27, 2026)**: Anthropic is running a 2-week promotion with **2x usage limits outside peak hours**. Peak hours are 5-11am PT (8am-2pm ET). All other hours — including evenings, nights, early mornings — get doubled usage. Weekends are fully 2x all day. Bonus usage does **not** count against your weekly plan limits. Applies to Free, Pro, Max, and Team plans (Enterprise excluded). Source: [Anthropic support article](https://support.claude.com/en/articles/14063676-claude-march-2026-usage-promotion).
>
> **For European users**: 5-11am PT = 13h-19h CET (France). So 2x applies from midnight to 1pm and from 7pm to midnight, plus all weekends.
**How Subscription Limits Work**
Unlike API usage (pay-per-token), subscriptions use a hybrid model that's deliberately opaque:
@ -11363,6 +11367,88 @@ git clone https://github.com/kairn-ai/kairn && cd kairn && pip install -e .
> **Source**: [kairn-ai/kairn GitHub](https://github.com/kairn-ai/kairn) (MIT, Python 100%)
### ICM: Dual Memory Architecture (Rust Binary, Zero Dependencies)
> **⚠️ Status: Under Testing** — Evaluated March 2026. Source-Available license (free for individuals and teams ≤20). From the rtk-ai team (same authors as RTK). Benchmarks below are vendor-reported and unverified independently. Feedback welcome!
**Purpose**: Persistent memory for AI agents combining episodic decay (Memories) and permanent knowledge graph (Memoirs) in a single zero-dependency Rust binary.
**When ICM makes sense over Kairn/doobidoo**:
- Python dependency management is a friction point (CI environments, sandboxed machines)
- You want Homebrew install with no Python env setup
- You need both decay-based episodic memory and a permanent knowledge graph in one tool
- You use multiple editors (14 clients supported: Claude Code, Cursor, VS Code, Windsurf, Zed, Amp, Cline, Roo Code, OpenAI Codex CLI, and more)
**Key differentiators vs Kairn/doobidoo**:
- **Single Rust binary**: no Python, no pip, no virtual env — `brew install icm` and done
- **Dual architecture in one tool**: Memories (decay, episodic) + Memoirs (permanent, typed graph) — Kairn covers the graph layer, doobidoo the semantic layer, ICM covers both
- **Auto-extraction**: three-layer automatic capture (pattern hooks, pre-compaction, session-start) without explicit `store_memory` calls
- **Auto-deduplication**: blocks entries with >85% similarity to existing content
| Feature | doobidoo | Kairn | ICM |
|---------|----------|-------|-----|
| Language | Python | Python | Rust (single binary) |
| Install | pip | pip | Homebrew / curl |
| Episodic decay | No | Yes (biological) | Yes (configurable rates) |
| Permanent knowledge graph | No | Yes | Yes (Memoirs) |
| Auto-extraction | No | No | Yes (3 layers) |
| Hybrid search | Semantic | Full-text + semantic | BM25 30% + vector 70% |
| License | MIT | MIT | Source-Available |
**Memoir relation types** (9): `part_of`, `depends_on`, `related_to`, `contradicts`, `refines`, `alternative_to`, `caused_by`, `instance_of`, `superseded_by`
**Installation**:
```bash
# Homebrew (recommended)
brew tap rtk-ai/tap && brew install icm
# Quick install
curl -fsSL https://raw.githubusercontent.com/rtk-ai/icm/main/install.sh | sh
# From source
cargo install --path crates/icm-cli
```
**MCP Config** (auto-configured via `icm init`):
```bash
icm init # interactive setup — detects and configures your editor
```
**Usage**:
```bash
# Store episodic memory (auto-decay)
icm store -t "my-project" -c "Use PostgreSQL for main DB" -i high
# Recall with hybrid search
icm recall "database choice" --topic "my-project"
# Build permanent knowledge graph
icm memoir create -n "system-architecture"
icm memoir add-concept -m "system-architecture" -n "auth-service"
icm memoir link -m "system-architecture" --from "api-gateway" --to "auth-service" -r depends-on
```
**Performance** (1000 ops, 384d embeddings — vendor-reported):
| Operation | Latency |
|-----------|---------|
| Store (no embeddings) | 34.2 µs/op |
| Store (with embeddings) | 51.6 µs/op |
| FTS5 full-text search | 46.6 µs/op |
| Vector search (KNN) | 590.0 µs/op |
| Hybrid search | 951.1 µs/op |
**Agent efficiency claims** (vendor-reported, Haiku model, unverified independently):
- Session 2: 29% fewer turns, 17% cost reduction
- Session 3: 40% fewer turns, 22% cost reduction
> ⚠️ **License note**: Free for individuals and teams of up to 20 people. Enterprise license required above that threshold. Verify your organization's size before deploying. Contact: license@rtk.ai
> **Source**: [rtk-ai/icm GitHub](https://github.com/rtk-ai/icm) (52 stars, Source-Available)
### MCP Memory Stack: Complementarity Patterns
> **⚠️ Experimental** - These patterns combine multiple MCP servers. Test in your workflow before relying on them.
@ -11389,15 +11475,18 @@ git clone https://github.com/kairn-ai/kairn && cd kairn && pip install -e .
**Comparison Matrix**:
| Capability | Serena | grepai | doobidoo | Kairn |
|------------|--------|--------|----------|-------|
| Cross-session memory | Key-value | No | Semantic | Knowledge graph |
| Cross-IDE memory | No | No | Yes | Yes |
| Cross-device sync | No | No | Yes (Cloudflare) | No |
| Knowledge Graph | No | Call graph | Decision graph | Typed relationships |
| Fuzzy search | No | Code | Memory | Full-text + semantic |
| Tags/categories | No | No | Yes | Yes |
| Memory decay / auto-expiry | No | No | No | Yes (biological) |
| Capability | Serena | grepai | doobidoo | Kairn | ICM |
|------------|--------|--------|----------|-------|-----|
| Cross-session memory | Key-value | No | Semantic | Knowledge graph | Episodic + graph |
| Cross-IDE memory | No | No | Yes | Yes | Yes (14 clients) |
| Cross-device sync | No | No | Yes (Cloudflare) | No | No |
| Knowledge Graph | No | Call graph | Decision graph | Typed relationships | Typed relationships |
| Fuzzy search | No | Code | Memory | Full-text + semantic | BM25 + vector hybrid |
| Tags/categories | No | No | Yes | Yes | Yes (topics) |
| Memory decay / auto-expiry | No | No | No | Yes (biological) | Yes (configurable) |
| Auto-extraction | No | No | No | No | Yes (3 layers) |
| Runtime | — | — | Python | Python | Rust (single binary) |
| License | MIT | MIT | MIT | MIT | Source-Available (≤20 free) |
**Usage Patterns**: