Daniel Chyan

[Daniel Chyan @Daniel_Chyan]

Surprised by glm-5.2 benchmark results

[Daniel Chyan @Daniel_Chyan]

Underrated skill with Fable 5: self-censorship

[Daniel Chyan @Daniel_Chyan]

Fable 5 on ultracode is very thorough

[Daniel Chyan @Daniel_Chyan]

@tanpukunokami Yes. Either as long as it’s “smooth” peanut butter

[Daniel Chyan @Daniel_Chyan]

gemini 3 flash is underrated. 1m context and price competitive with open models

[Daniel Chyan @Daniel_Chyan]

Leaderboard: https://t.co/0HTxPgTCYG

Repo: https://t.co/ZhFyxIgDPY

[Daniel Chyan @Daniel_Chyan]

Most agent evals are over in minutes.

I wanted one that tests whether agents can stay useful for hours, adapt through phase changes, and recover from bad strategy.

So I made a Harbor benchmark where agents play Universal Paperclips until they convert the universe into
paperclips.

No source access, hidden state, or JS eval. Best run: GPT-5.5 xhigh, 7h39m.

[Daniel Chyan @Daniel_Chyan]

@innoutburger_ Light well is the best

[Daniel Chyan @Daniel_Chyan]

Thoroughly enjoying gpt-5.5 in chatgpt

Try out Thinking + Extended. Great brainstorming partner

[Daniel Chyan @Daniel_Chyan]

@OpenRouter is great future proofing. Can A/B test opus 4.7, gpt 5.5, and GLM 5.1

[Daniel Chyan @Daniel_Chyan]

ChatGPT 5.5 Thinking (Extended) and the new image generation model cooks.

I wanted to caramelize onions faster. Prompted chem lab techniques. It delivered.

[Daniel Chyan @Daniel_Chyan]

@andrewchen Unfortunate for LA

[Daniel Chyan @Daniel_Chyan]

Why Opus 4.7 Adaptive is way better in Anthropic's apps than in Openclaw's harness. Openclaw's system prompts are quite harsh.
https://t.co/KwPev08LO2

[Daniel Chyan @Daniel_Chyan]

Use Opus 4.7 on high or higher. Adaptive takes you down misleading rabbit holes and enraging debugging sessions that cost more tokens than starting on high+ in the first place.

[Daniel Chyan @Daniel_Chyan]

@innoutburger_ Animal style remains the same- whew

[Daniel Chyan @Daniel_Chyan]

type a vibe. get a lo-fi track with a movie quote built in.
built notbumblebee for ElevenHacks @turbopuffer x @ElevenLabs : hybrid vector search finds the perfect dialogue from 10K movie clips, then ElevenLabs Music API composes a full track around it.

[Daniel Chyan @Daniel_Chyan]

@annimaniac Glad to hear we’re in good company! Slight variation from us: Gemini is our review agent

[Daniel Chyan @Daniel_Chyan]

@PlotWeaver Dependent on population density. In the city, 0 is most common. Suburbs 1-2. Rural, as many as you can afford.

Enthusiast would be 3: handgun, rifle, shotgun

[Daniel Chyan @Daniel_Chyan]

Who runs https://t.co/4BrJuScORy?

I want to submit a successful 9.5h run with claude code sonnet.

[Daniel Chyan @Daniel_Chyan]

@agentmail My agent is grateful https://t.co/gwWquS7PRu