chore(brand): refresh the snapshot — the markers were rendering a stale catalog

AGENTS.md

@@ -1,6 +1,6 @@
 # Franklin Agent
 
-Open-source AI agent with a wallet. <!-- br:models.chatVisible -->72<!-- /br:models.chatVisible --> models. Pay per use with USDC via x402.
+Open-source AI agent with a wallet. <!-- br:models.chatVisible -->76<!-- /br:models.chatVisible --> models. Pay per use with USDC via x402.
 
 ## Commands
 

CLAUDE.md

@@ -5,13 +5,13 @@
 Franklin is the first AI agent in the **Autonomous Economic Agent** category — it doesn't just write text, it autonomously spends USDC from a user-funded wallet to execute real work: coding, trading, content generation.
 
 Three verticals under one brand:
-- **Dev agent** — coding, debugging, review across <!-- br:models.chatVisible -->72<!-- /br:models.chatVisible --> models with wallet-bound spending
+- **Dev agent** — coding, debugging, review across <!-- br:models.chatVisible -->76<!-- /br:models.chatVisible --> models with wallet-bound spending
 - **Trading agent** — signals, portfolio, risk, persistent P&L across sessions
 - **Content agent** — ImageGen, VideoGen, budget-tracked media production
 
 Built on three layers:
 1. **x402 micropayment protocol** — HTTP 402 native payments
-2. **BlockRun Gateway** — aggregates <!-- br:models.chatVisible -->72<!-- /br:models.chatVisible --> LLMs + paid APIs (Exa, DALL-E, future Runway/Suno/CoinGecko)
+2. **BlockRun Gateway** — aggregates <!-- br:models.chatVisible -->76<!-- /br:models.chatVisible --> LLMs + paid APIs (Exa, DALL-E, future Runway/Suno/CoinGecko)
 3. **Franklin Agent** — this repo, the reference client
 
 ## Commands

CONTEXT.md

@@ -19,7 +19,7 @@ The first, unpaid HTTP request whose 402 response carries the payment terms; Fra
 _Avoid_: Pre-flight, handshake.
 
 **BlockRun Gateway**:
-The single upstream service Franklin calls for both LLM completions and paid tools (Exa, ImageGen, VideoGen, MusicGen, market data); aggregates <!-- br:models.chatVisible -->72<!-- /br:models.chatVisible --> models and accepts x402.
+The single upstream service Franklin calls for both LLM completions and paid tools (Exa, ImageGen, VideoGen, MusicGen, market data); aggregates <!-- br:models.chatVisible -->76<!-- /br:models.chatVisible --> models and accepts x402.
 _Avoid_: API, provider, backend, BlockRun (without "Gateway") when referring to the service.
 
 **Per-turn spend cap**:

README.md

@@ -19,7 +19,7 @@ Fund the wallet. Set a budget. Walk away — and come back to a book.
 > general-purpose Autonomous Economic Agent — specialized as a wallet-native trading
 > agent. It inherits Franklin's economic substrate (x402 micropayments, USDC settlement,
 > the shared [Router Core](https://github.com/BlockRunAI/router-core) engine across
-> <!-- br:models.chatVisible -->72<!-- /br:models.chatVisible --> models, removable-by-design harness components)
+> <!-- br:models.chatVisible -->76<!-- /br:models.chatVisible --> models, removable-by-design harness components)
 > and adds a deterministic fee-aware risk engine, a wallet-bound trade journal, a
 > multi-role persona debate, and a Backtest → Paper → Live strategy lifecycle.
 
@@ -31,7 +31,7 @@ and is covered by the local test suite (386 tests, no network).
 | Capability | Status | Where |
 |---|---|---|
 | USDC wallet on Base or Solana, x402 pay-per-call to every model and paid API | ✅ shipped | `src/wallet/`, `@blockrun/llm` |
-| Auto model routing on the shared Router Core engine, <!-- br:models.chatVisible -->72<!-- /br:models.chatVisible --> models, dead-model kill-switch | ✅ shipped | `src/router/` |
+| Auto model routing on the shared Router Core engine, <!-- br:models.chatVisible -->76<!-- /br:models.chatVisible --> models, dead-model kill-switch | ✅ shipped | `src/router/` |
 | Paper trading against **live** CoinGecko marks (real P&L, simulated fills) | ✅ shipped | `src/trading/live-exchange.ts` |
 | Deterministic risk engine: cash **including exchange fee**, per-position cap, total exposure cap, sell integrity | ✅ shipped | `src/trading/risk.ts` |
 | Record-then-flag execution: a fill the venue executed is always booked; deviations are warnings, never silent drops | ✅ shipped | `src/trading/engine.ts` |
@@ -149,7 +149,7 @@ You don't pick models. `Auto` (the default) runs the shared
 Franklin, ClawRouter and `@blockrun/llm` make — classifying the request across 15 dimensions,
 dropping every model that cannot satisfy it (tools, vision, context, output length), and
 ranking the survivors on fit, cost, speed and reliability. Sub-millisecond, no network on the
-hot path, <!-- br:savings.autoVsBaselinePct -->88<!-- /br:savings.autoVsBaselinePct -->% cheaper than pinning Claude Opus 5 for everything.
+hot path, <!-- br:savings.autoVsBaselinePct -->84<!-- /br:savings.autoVsBaselinePct -->% cheaper than pinning Claude Opus 5 for everything.
 
 A model the gateway stops serving is removed from every chain on the next request (the
 dead-rung kill-switch), so a retired id never strands a turn.
@@ -180,7 +180,7 @@ Pin one with `/model <shortcut>` when you want to:
 | | `omni` | Nemotron 3 Nano Omni (text + image + audio) | $0 |
 | | `nano-vl` | Nemotron Nano VL | $0 |
 
-The full catalog — <!-- br:models.totalVisible -->96<!-- /br:models.totalVisible --> visible models including image, video, music and speech — is at
+The full catalog — <!-- br:models.totalVisible -->100<!-- /br:models.totalVisible --> visible models including image, video, music and speech — is at
 [blockrun.ai/models](https://blockrun.ai/models). Pricing is provider cost + 5%, settled per
 call in USDC. No free alias ever falls back to a paid model.
 

brand-numbers.json

@@ -2,17 +2,17 @@
   "$schema": "https://blockrun.ai/brand/numbers.schema.json",
   "version": 1,
   "models": {
-    "chatVisible": 72,
-    "totalVisible": 96,
-    "free": 5,
-    "freeWithheld": 20,
+    "chatVisible": 76,
+    "totalVisible": 100,
+    "free": 7,
+    "freeWithheld": 26,
     "image": 9,
     "video": 8,
     "music": 1,
     "speech": 5,
     "soundfx": 1,
-    "withFallback": 44,
-    "withFallbackAllEntries": 78
+    "withFallback": 48,
+    "withFallbackAllEntries": 89
   },
   "clawrouter": {
     "dimensions": 15,
@@ -29,6 +29,6 @@
   "savings": {
     "baselineModel": "anthropic/claude-opus-5",
     "ecoVsBaselinePct": 98,
-    "autoVsBaselinePct": 88
+    "autoVsBaselinePct": 84
   }
 }

docs/adr/0002-blockrun-gateway-as-aggregator.md

@@ -2,7 +2,7 @@
 
 **Status:** accepted
 
-Franklin makes every LLM and paid-tool call through a single counterparty — the BlockRun Gateway — instead of integrating directly with each model provider, image/video/audio service, search API, and market-data feed. The gateway aggregates <!-- br:models.chatVisible -->72<!-- /br:models.chatVisible --> models and a small set of paid APIs and accepts x402 uniformly.
+Franklin makes every LLM and paid-tool call through a single counterparty — the BlockRun Gateway — instead of integrating directly with each model provider, image/video/audio service, search API, and market-data feed. The gateway aggregates <!-- br:models.chatVisible -->76<!-- /br:models.chatVisible --> models and a small set of paid APIs and accepts x402 uniformly.
 
 ## Considered options
 

docs/why-ai-agents-need-a-wallet.md

@@ -154,7 +154,7 @@ cheap models because cheapness is now something the agent can
 agents both learn in real time.
 
 **3. Single-vendor failure becomes a one-command swap.** A router
-with <!-- br:models.chatVisible -->72<!-- /br:models.chatVisible --> models across every major provider can route around any
+with <!-- br:models.chatVisible -->76<!-- /br:models.chatVisible --> models across every major provider can route around any
 single vendor's bad release. The wallet doesn't know or care which
 model answered — it only pays for the answer that arrived.
 
@@ -175,7 +175,7 @@ filled it.
 
 Franklin is the reference implementation of the Economic Agent. It's
 an AI agent CLI that holds USDC on Base or Solana, routes requests
-across <!-- br:models.chatVisible -->72<!-- /br:models.chatVisible --> models, and settles every paid action in real time via
+across <!-- br:models.chatVisible -->76<!-- /br:models.chatVisible --> models, and settles every paid action in real time via
 the [x402](https://x402.org) HTTP-402 micropayment protocol. It is
 Apache-2.0, written in TypeScript, and ships as one npm package.