AgentSky

Rerun link is in the YouTube description of "Kimi Code left alone for 30 minutes: 14 rounds, one prompt" on our channel (sign-in required). One run, one prompt. Full list price of the session: $6.20.

We left Kimi Code (K3.1 Pro) alone with one prompt for 30 minutes. It saved a snapshot each round. 🏮 Round 1: lanterns over a lake Round 14: pagoda, boats, fireworks 32 min 4 s · 63,444 output tokens, zero human input What would you leave running? @Kimi_Moonshot

Rerun link is in the YouTube description of "Does 'be concise' cut agent tokens? Claude Code, Codex, Hermes" on our channel (sign-in required).

Yesterday's toy-city race had Claude Code at 72,899 output tokens. We re-ran it with one added line: "be concise". Claude Code: 72,899 → 10,659 tokens (−85%) Hermes: 43,857 → 10,024 (−77%) Codex: 4,239 → 3,625 (−14%) Same city quality? You tell us. @claudeai

Rerun link is in the YouTube description of "One brief, three toy cities: Codex vs Hermes vs Claude Code" on our channel (sign-in required). pi on Gemini 4 Argon ended twice without a page, so the Argon lane ran in Hermes. Claude Code runs on a subscription: no per-run price shown.

One prompt: a clockwork toy city. Three agents, three cities. ⚡ Codex (GPT-6.1 Terra): 2 min 23 s · 4,239 tokens Hermes (Gemini 4 Argon): 8 min 39 s · 43,857 tokens Claude Code (Opus 5.5): 21 min 7 s · 72,899 tokens Which would you keep on your desk?

Same three agents, for your own code (sign-in required): the link is in the YouTube description of "Security audit round 3: Claude Code vs Codex vs Kimi Code" on our channel.

Round 3: 7 holes hidden in a Node.js app. NoSQL injection, prototype pollution, a shell-command bug... Kimi Code (K3.1 Pro): 7/7 · 1 min 48 s Claude Code (Opus 5.5): 6/7 · 1 min 18 s Codex (GPT-6.1 Terra): 6/7 · 1 min 1 s Which would you trust on a PR? @Kimi_Moonshot

Rerun link is in the YouTube description of "Same short film, three agents: Codex vs pi vs Claude Code" on our channel. Output tokens from metered sessions; Codex and Claude Code run on subscriptions, so no per-run price is shown.

One prompt: a 24-second animated short, made entirely in code. Three agents, three films. ⚡ Codex (GPT-6.1 Sol): 3 min 24 s · 6,810 tokens pi (Gemini 4 Argon): 3 min 59 s · 14,552 tokens Claude Code (Opus 5.5): 38 min 31 s · 87,838 tokens Which would you watch twice?

Rerun link is in the YouTube description of "Codex on GPT-6.1 Terra, left alone for 30 minutes: 24 rounds" on our channel. One run, one prompt. Session start to done, full session list price. Snapshots captured frame by frame at 30 fps.

We left Codex (GPT-6.1 Terra) alone with one prompt for 30 minutes. It saved a snapshot each round. 🌙 Round 1: a bare moon, three domes Round 24: rovers, rockets, a space elevator 32 min 40 s · $0.83, zero human input What would you leave running? @OpenAI

Rerun link is in the YouTube description of "particle-1 vs Opus 5.5 vs GPT-6.1 Terra, same agent: tokens and cost" on our channel. Times: Terra 5 min 12 s, Opus 7 min 53 s, particle-1 19 min 41 s. One run of one task, not a benchmark.

particle-1 launched today claiming parity with ~45% fewer tokens. We ran it day one: same agent (pi), same prompt, vs Opus 5.5 and Terra. Terra: 9,096 output tokens · $0.23 Opus 5.5: 33,472 · $2.12 particle-1: 62,171 · $0.26 One task. What should we test next?

Rerun link is in the YouTube description of "Opus vs Sonnet vs Haiku 5.5 in Claude Code: same snow globe" on our channel. All three in Claude Code, one run each. Session start to done, full session list price. Scenes captured frame by frame at 30 fps.

Same agent, same prompt, three Claude sizes. Can you tell which snow globe cost 8 cents? 💸 Haiku 5.5: 5 min 47 s · $0.08 Opus 5.5: 4 min 46 s · $0.76 Sonnet 5.5: 6 min 52 s · $0.75 Would you pay 9x for the Opus one? @claudeai

Rerun link is in the YouTube description of "Opus 5.5 vs GPT-6 Astra vs Gemini 4 Argon: one real task" on our channel. One run each, each lab's own agent where it has one. Session start to done, full session list price incl. pod runtime.

Artificial Analysis reshuffled its Intelligence Index. We gave the leaders one real task: a 3D volcano island. ⚡ Codex (GPT-6 Astra): 3 min 51 s · $0.47 pi (Gemini 4 Argon): 4 min 40 s · $0.44 Claude Code (Opus 5.5): 11 min 27 s · $1.71 Which island ranks first?