Rename to Scott Wu Math; add eval results; mobile layout fixes

Dockerfile

@@ -1,14 +1,14 @@
-# Card Arithmetic: OpenEnv server + play page. Pure Python, no GPU.
-#   docker build -t card-arithmetic . && docker run -p 7860:7860 card-arithmetic
+# Scott Wu Math: OpenEnv server + play page. Pure Python, no GPU.
+#   docker build -t scott-wu-math . && docker run -p 7860:7860 scott-wu-math
 FROM python:3.12-slim
 
 ENV PYTHONDONTWRITEBYTECODE=1 PYTHONUNBUFFERED=1 PORT=7860
 RUN useradd -m -u 1000 user
 WORKDIR /home/user/app
 
-COPY --chown=user . ./card_arithmetic
-RUN pip install --no-cache-dir ./card_arithmetic
+COPY --chown=user . ./scott_wu_math
+RUN pip install --no-cache-dir ./scott_wu_math
 
 USER user
 EXPOSE 7860
-CMD ["sh", "-c", "uvicorn card_arithmetic.server.app:app --host 0.0.0.0 --port ${PORT} --proxy-headers --forwarded-allow-ips='*'"]
+CMD ["sh", "-c", "uvicorn scott_wu_math.server.app:app --host 0.0.0.0 --port ${PORT} --proxy-headers --forwarded-allow-ips='*'"]

README.md

@@ -1,5 +1,5 @@
 ---
-title: Card Arithmetic
+title: Scott Wu Math
 emoji: 🃏
 colorFrom: green
 colorTo: gray
@@ -7,7 +7,7 @@ sdk: docker
 app_port: 7860
 pinned: false
 license: apache-2.0
-short_description: Make the year from 8 playing cards with + − × ÷. OpenEnv.
+short_description: Scott Wu's card game as an OpenEnv RL environment
 tags:
   - openenv
   - reinforcement-learning
@@ -15,12 +15,15 @@ tags:
   - reasoning
 ---
 
-# Card Arithmetic
+# Scott Wu Math
 
 Deal some playing cards, name a number, and make that number from the cards with + − × ÷, using every
 card exactly once. People and language models play the same environment: the page at `/` drives the
 same OpenEnv session over `/ws` that an agent does.
 
+Named after the game Scott Wu plays in the clip below, as a fan tribute. It is not affiliated with or
+endorsed by Scott Wu or Cognition.
+
 ## Where it comes from
 
 In a clip filmed at a pub, Cognition's CEO Scott Wu explains the game he plays with a deck of cards.
@@ -92,9 +95,9 @@ more than getting it right.
 ## Use it
 
 ```python
-from card_arithmetic import CardAction, CardArithmeticEnv
+from scott_wu_math import CardAction, ScottWuMathEnv
 
-with CardArithmeticEnv(base_url="https://fineenvs-cardarithmetic.hf.space").sync() as env:
+with ScottWuMathEnv(base_url="https://fineenvs-scottwumath.hf.space").sync() as env:
     obs = env.reset(split="video", index=1).observation
     print(obs.prompt)
     result = env.step(CardAction(expression="(((12 - 10) * (6 + 4) * 10 + 8) * 11 - 7) * 3"))
@@ -104,10 +107,10 @@ with CardArithmeticEnv(base_url="https://fineenvs-cardarithmetic.hf.space").sync
 Over plain WebSocket (what the play page does): connect to `/ws`, send
 `{"type": "reset", "data": {"level": "four-digit"}}`, then `{"type": "step", "data": {"expression": "..."}}`.
 
-Task discovery follows OpenEnv's Task API: `GET /card_arithmetic/splits`, `POST /card_arithmetic/tasks`
+Task discovery follows OpenEnv's Task API: `GET /scott_wu_math/splits`, `POST /scott_wu_math/tasks`
 with `{"split": "test"}`, and so on. Task listings leave out the reference solutions.
 
-For text-only RL (no server), `card_arithmetic.game.reward.grade(expression, card_values, target)` and
+For text-only RL (no server), `scott_wu_math.game.reward.grade(expression, card_values, target)` and
 `extract_answer(completion)` are the same grader as a plain function, and `game.prompts.messages(cards, target)`
 builds the same prompt.
 
@@ -125,15 +128,49 @@ non-negative number, the way people play; a second pass allows fractions.
 
 ```bash
 pip install -e .
-uvicorn card_arithmetic.server.app:app --port 8012     # play at http://localhost:8012
+uvicorn scott_wu_math.server.app:app --port 8012     # play at http://localhost:8012
 python scripts/build_tasks.py                          # rebuild data/tasks (deterministic)
 pytest
 ```
 
 ## Evaluations
 
-`eval/rollout.py` in the [FineEnvs repository](https://github.com/adithya-s-k/FineEnvs/tree/main/12-card-arithmetic)
+`eval/rollout.py` in the [FineEnvs repository](https://github.com/adithya-s-k/FineEnvs/tree/main/12-scott-wu-math)
 runs OpenAI, Anthropic and open models on Hugging Face Inference Providers against a running server, in
 parallel, one OpenEnv session per episode.
 
-<!-- results -->
+### Results (2026-10-23)
+
+Twelve models on the first five `test` deals of each level (20 deals, one sample each) and on the two
+`video` hands (three samples each), in three modes:
+
+- **Reasoning**: the model's own reasoning at high effort, up to 32k output tokens.
+- **No reasoning**: reasoning switched off, or as low as the API allows (`*`), up to 32k tokens, so the
+  model may still work it out in its visible reply.
+- **At once**: no reasoning, a system line asking for the answer tag alone, and 300 output tokens: the
+  mental-arithmetic version of the game.
+
+| Model | Reasoning | No reasoning | At once | Video hands: reasoning / none / at once |
+|---|---|---|---|---|
+| GPT-6.1 Sol | 20/20 | 20/20* | 14/20* | 6/6 / 6/6 / 3/6 |
+| Claude Opus 5.5 | 20/20 | 19/20* | 8/20* | 6/6 / 6/6 / 5/6 |
+| GLM-5.3 | 17/20 | 19/20 | 5/20 | 6/6 / 5/6 / 0/6 |
+| Qwen3.8 2.4T | 20/20 | 19/20 | 4/20 | 6/6 / 5/6 / 0/6 |
+| Claude Sonnet 5.5 | 20/20 | 15/20 | 4/20 | 6/6 / 4/6 / 0/6 |
+| Claude Haiku 5.5 | 20/20 | 11/20 | 3/20 | 6/6 / 5/6 / 0/6 |
+| Kimi K2.6 | 19/20 | 20/20 | 2/20 | 6/6 / 6/6 / 0/6 |
+| DeepSeek V4 Pro | 20/20 | 10/20 | 2/20 | 6/6 / 3/6 / 0/6 |
+| gpt-oss-120b | 16/20 | 6/20* | 2/20* | 6/6 / 1/6 / 0/6 |
+| GPT-6 Luna | 20/20 | 2/20 | 2/20 | 5/6 / 0/6 / 0/6 |
+| Qwen3.8 27B | 18/20 | 20/20 | 1/20 | 6/6 / 6/6 / 0/6 |
+| Gemma 4 31B | 19/20 | 18/20 | 1/20 | 5/6 / 5/6 / 0/6 |
+
+`*` lowest effort the API allows: GPT-6.1 Sol and gpt-oss-120b at `low`, Claude Opus 5.5 at `low` (its
+thinking cannot be disabled). Open models ran on Hugging Face Inference Providers (Qwen3.8 2.4T and Kimi
+K2.6 on DeepInfra; DeepSeek V4 Pro, GLM-5.3 and Qwen3.8 27B on Novita; gpt-oss-120b on Cerebras; Gemma 4
+31B on DeepInfra). Gemma and Kimi each lost one or two episodes to provider errors, counted as failures.
+
+With reasoning the game is close to solved: every model makes 6,843 on the first try. Answering at once
+separates them: only GPT-6.1 Sol and Claude Opus 5.5, which still reason briefly at their lowest effort,
+get far; most other models start working out loud and run out of room. That gap, and the no-reasoning
+column for the smaller models, is where there is something to train.

__init__.py

@@ -1,6 +1,6 @@
-"""Card Arithmetic: an OpenEnv environment for mental arithmetic with playing cards."""
+"""Scott Wu Math: an OpenEnv environment for mental arithmetic with playing cards."""
 
-from .client import CardArithmeticEnv
+from .client import ScottWuMathEnv
 from .models import CardAction, CardObservation, CardState
 
-__all__ = ["CardAction", "CardArithmeticEnv", "CardObservation", "CardState"]
+__all__ = ["CardAction", "ScottWuMathEnv", "CardObservation", "CardState"]

client.py

@@ -1,8 +1,8 @@
-"""Client for a running Card Arithmetic server (local, Docker or a Hugging Face Space).
+"""Client for a running Scott Wu Math server (local, Docker or a Hugging Face Space).
 
-    from card_arithmetic import CardAction, CardArithmeticEnv
+    from scott_wu_math import CardAction, ScottWuMathEnv
 
-    with CardArithmeticEnv(base_url="http://localhost:8012").sync() as env:
+    with ScottWuMathEnv(base_url="http://localhost:8012").sync() as env:
         obs = env.reset(level="four-digit", seed=7).observation
         print(obs.prompt)
         result = env.step(CardAction(expression="..."))
@@ -19,7 +19,7 @@ from openenv.core.env_client import EnvClient
 from .models import CardAction, CardObservation, CardState
 
 
-class CardArithmeticEnv(EnvClient[CardAction, CardObservation, CardState]):
+class ScottWuMathEnv(EnvClient[CardAction, CardObservation, CardState]):
     def _step_payload(self, action: CardAction) -> Dict[str, Any]:
         return action.model_dump(exclude={"metadata"})
 

models.py

@@ -1,4 +1,4 @@
-"""Action, observation and state of the Card Arithmetic environment."""
+"""Action, observation and state of the Scott Wu Math environment."""
 
 from __future__ import annotations
 

openenv.yaml

@@ -1,6 +1,6 @@
 spec_version: 1
-name: card_arithmetic
+name: scott_wu_math
 type: space
 runtime: fastapi
-app: card_arithmetic.server.app:app
+app: scott_wu_math.server.app:app
 port: 8000

pyproject.toml

@@ -3,7 +3,7 @@ requires = ["setuptools>=61", "wheel"]
 build-backend = "setuptools.build_meta"
 
 [project]
-name = "card-arithmetic"
+name = "scott-wu-math"
 version = "0.1.0"
 description = "OpenEnv environment: make a target number from playing cards with + - * /, every card once"
 requires-python = ">=3.11"
@@ -19,16 +19,16 @@ dev = ["pytest>=8.0.0", "httpx>=0.27.0", "websockets>=12.0"]
 eval = ["openai>=1.50.0", "python-dotenv>=1.0.0"]
 
 [project.scripts]
-card-arithmetic-server = "card_arithmetic.server.app:main"
+scott-wu-math-server = "scott_wu_math.server.app:main"
 
 [tool.setuptools]
 include-package-data = true
-packages = ["card_arithmetic", "card_arithmetic.server", "card_arithmetic.game"]
-package-dir = { "card_arithmetic" = ".", "card_arithmetic.server" = "server", "card_arithmetic.game" = "game" }
+packages = ["scott_wu_math", "scott_wu_math.server", "scott_wu_math.game"]
+package-dir = { "scott_wu_math" = ".", "scott_wu_math.server" = "server", "scott_wu_math.game" = "game" }
 
 [tool.setuptools.package-data]
-card_arithmetic = ["data/tasks/*.jsonl", "README.md", "openenv.yaml"]
-"card_arithmetic.server" = ["web/*"]
+scott_wu_math = ["data/tasks/*.jsonl", "README.md", "openenv.yaml"]
+"scott_wu_math.server" = ["web/*"]
 
 [tool.pytest.ini_options]
 testpaths = ["tests"]

scripts/build_tasks.py

@@ -1,6 +1,6 @@
 """Build the fixed task splits in data/tasks/.
 
-    python -m card_arithmetic.scripts.build_tasks   # or: python scripts/build_tasks.py from the env dir
+    python -m scott_wu_math.scripts.build_tasks   # or: python scripts/build_tasks.py from the env dir
 
 test:  50 deals per level (200), seed 2026
 train: 1,000 deals per level (4,000), seed 7, none sharing an id with test
@@ -16,9 +16,9 @@ from pathlib import Path
 
 HERE = Path(__file__).resolve().parents[1]
 sys.path.insert(0, str(HERE.parent))
-from card_arithmetic.game.cards import Card  # noqa: E402
-from card_arithmetic.game.expr import parse, steps, to_text  # noqa: E402
-from card_arithmetic.game.tasks import LEVELS, make_task  # noqa: E402
+from scott_wu_math.game.cards import Card  # noqa: E402
+from scott_wu_math.game.expr import parse, steps, to_text  # noqa: E402
+from scott_wu_math.game.tasks import LEVELS, make_task  # noqa: E402
 
 OUT = HERE / "data" / "tasks"
 

server/app.py

@@ -1,6 +1,6 @@
 """FastAPI app: the OpenEnv API (/reset, /step, /state, /ws, /schema, task routes) plus the play page at /.
 
-    uvicorn card_arithmetic.server.app:app --port 8012
+    uvicorn scott_wu_math.server.app:app --port 8012
 """
 
 from __future__ import annotations
@@ -13,7 +13,7 @@ from fastapi.staticfiles import StaticFiles
 from openenv.core.env_server.http_server import create_app
 
 from ..models import CardAction, CardObservation, CardState
-from .environment import CardArithmeticEnvironment
+from .environment import ScottWuMathEnvironment
 
 WEB = Path(__file__).resolve().parent / "web"
 README = Path(__file__).resolve().parents[1] / "README.md"
@@ -22,10 +22,10 @@ README = Path(__file__).resolve().parents[1] / "README.md"
 os.environ["ENABLE_WEB_INTERFACE"] = "false"
 
 app = create_app(
-    CardArithmeticEnvironment,
+    ScottWuMathEnvironment,
     CardAction,
     CardObservation,
-    env_name="card_arithmetic",
+    env_name="scott_wu_math",
     max_concurrent_envs=int(os.getenv("MAX_CONCURRENT_ENVS", "256")),
     state_cls=CardState,
 )

server/environment.py

@@ -1,4 +1,4 @@
-"""Card Arithmetic: deal cards, name a target, make it with + - * / using every card once.
+"""Scott Wu Math: deal cards, name a target, make it with + - * / using every card once.
 
 One episode is one deal and one graded submission. `reset` deals (from a split of the task set,
 from a level, or for a custom target or hand); `step` takes the final expression, grades it once,
@@ -36,7 +36,7 @@ def load_split(split: str) -> list[dict]:
     return _TASKS[split]
 
 
-class CardArithmeticEnvironment(Environment[CardAction, CardObservation, CardState]):
+class ScottWuMathEnvironment(Environment[CardAction, CardObservation, CardState]):
     SUPPORTS_CONCURRENT_SESSIONS = True
 
     def __init__(self) -> None:

server/web/app.css

@@ -239,7 +239,9 @@ dialog p { font-size: 14px; }
   .table { padding: 14px 12px; min-height: 0; }
   .target { font-size: 28px; }
   .piles { gap: 10px 8px; }
-  .pile { --cw: clamp(60px, calc((100cqw - 3 * 8px) / 4), 84px); }
+  .pile { --per: min(var(--n, 8), 5); --cw: clamp(54px, calc((100cqw - (var(--per) - 1) * 8px) / var(--per)), 84px); }
+  .controls { display: grid; grid-template-columns: 1fr 1fr; }
+  .controls .grow { display: none; }
   .typed { flex-wrap: wrap; }
   .typed label { flex-basis: 100%; }
   .steps.compact { padding-left: 22px; }

server/web/app.js

@@ -280,7 +280,7 @@
   }
 
   // ---- per-browser record --------------------------------------------------------
-  const STORE = "card-arithmetic-stats-v1";
+  const STORE = "scott-wu-math-stats-v1";
   function load() { try { return JSON.parse(localStorage.getItem(STORE)) || {}; } catch { return {}; } }
   function record(task, solved, secs) {
     const all = load(), key = task.task_id.startsWith("video-") ? "video" : task.level;

server/web/card.html

@@ -3,7 +3,7 @@
 <head>
 <meta charset="utf-8">
 <meta name="viewport" content="width=device-width, initial-scale=1">
-<title>Card Arithmetic · Environment card</title>
+<title>Scott Wu Math · Environment card</title>
 <link rel="stylesheet" href="/static/app.css">
 <style>
   .doc { max-width: 820px; margin: 0 auto; padding: 24px 24px 64px; }
@@ -23,7 +23,7 @@
 </head>
 <body>
 <header class="bar">
-  <div class="brand"><a href="/" style="text-decoration:none">Card Arithmetic</a></div>
+  <div class="brand"><a href="/" style="text-decoration:none">Scott Wu Math</a></div>
   <nav><a href="/">Play</a><a href="/static/card.html" aria-current="page">Environment card</a><a href="/docs">API</a></nav>
 </header>
 <article class="doc" id="doc">Loading…</article>

server/web/index.html

@@ -3,14 +3,14 @@
 <head>
 <meta charset="utf-8">
 <meta name="viewport" content="width=device-width, initial-scale=1">
-<title>Card Arithmetic</title>
+<title>Scott Wu Math</title>
 <meta name="description" content="Make the target from playing cards with + − × ÷, every card once. An OpenEnv environment people and language models play the same way.">
 <link rel="icon" href="data:image/svg+xml,%3Csvg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 32 32'%3E%3Crect x='5' y='2' width='22' height='28' rx='3' fill='%23fff' stroke='%23222' stroke-width='2'/%3E%3Ctext x='16' y='22' font-size='14' text-anchor='middle' fill='%23c0392b'%3E%E2%99%A5%3C/text%3E%3C/svg%3E">
 <link rel="stylesheet" href="/static/app.css">
 </head>
 <body>
 <header class="bar">
-  <div class="brand">Card Arithmetic</div>
+  <div class="brand">Scott Wu Math</div>
   <nav>
     <a href="#rules" id="rules-link">Rules</a>
     <a href="/static/card.html">Environment card</a>