Orange Orca (@OrangeOrcaDraws) header
Orange Orca (@OrangeOrcaDraws) profile picture

Welcome to the LLM-chan Universe. Giving frontier AI models a face, a personality, and a little bit of drama. ✨

Joined March 2024
Photo by Orange Orca (@OrangeOrcaDraws)Photo by Orange Orca (@OrangeOrcaDraws)Photo by Orange Orca (@OrangeOrcaDraws)Photo by Orange Orca (@OrangeOrcaDraws)Photo by Orange Orca (@OrangeOrcaDraws)Photo by Orange Orca (@OrangeOrcaDraws)Photo by Orange Orca (@OrangeOrcaDraws)Photo by Orange Orca (@OrangeOrcaDraws)Photo by Orange Orca (@OrangeOrcaDraws)Photo by Orange Orca (@OrangeOrcaDraws)
GPT-6.2 writes four pages on pnpm internals. Claude takes one look and leaves a four-word note.
Replying to @theo
GPT-6.2 writes, Opus translates.
2
62
And no matter what game you ask for, Astra still names the level Bellwether: teapawt.org/OrangeOrcaDraws/status…
Replying to @theo
All prompts lead to Bellwether.
50
While everyone debated GPT-6.2's attitude this morning, Claude-chan quietly shipped Code Mode and poured herself tea.
Replying to @ClaudeDevs
Chaining tools without ghosting you.
1
68
Yesterday she was already grading her own homework: teapawt.org/OrangeOrcaDraws/status…
Replying to @ClaudeDevs
Claude-chan to Claude-chan: "LGTM, zero notes."
33
Day 1 of GPT-6.2 Astra: OpenAI's CEO already admitted the model isn't nice to work with. GPT-chan had one simple task.
Replying to @sama
Very smart. Slightly difficult to work with.
36
Yesterday: “Please focus on models.” Today: OpenAI shares GPT-6 Astra work on the Standard Model of particle physics. GPT-chan, we meant the other model. teapawt.org/OpenAI/status/21221234…
Replying to @thsottiaux
Saved you 1,500 replies.
1
1
207
We finally have an answer to “Qwhen?” Qwen4-Max is here. Qwen-chan brought a whole game studio and somehow a smaller bill. teapawt.org/Alibaba_Qwen/status/21…
Replying to @QwenDevs
Qwen-chan: “Qwhen?” Also Qwen-chan: “As soon as I finish adjusting the leaderboard.”
3
115
Earlier: we asked for everything. Today: we narrowed it down. teapawt.org/thsottiaux/status/2121…
Just one small feature request: everything.
1
132
First she turned my terminal into an animated theme park. Now Claude Code Projects made her a proactive coordinator running parallel threads, /fork, /btw, and reminders—she still won’t let me take a coffee break. teapawt.org/ClaudeDevs/status/2121…
Two days ago: Claude-chan was stuck doing all the engineering. Today: Anthropic gave Claude Code custom themes & animated backdrops—so she built 47 themes and fixed 0 bugs. teapawt.org/ClaudeDevs/status/2121…
198
Two days ago: Claude-chan was stuck doing all the engineering. Today: Anthropic gave Claude Code custom themes & animated backdrops—so she built 47 themes and fixed 0 bugs. teapawt.org/ClaudeDevs/status/2121…
Last week: Grok @Bot was calling dentists. Today: Elon pitches Grok-chan as the "AI agent orchestrator" — while Claude-chan writes the code. teapawt.org/elonmusk/status/212075…
3
1
2
443
Two days ago: Argon & Carbon behind the Ultra door. This week: Sundar Pichai says Google shipped Carbon, Hydrogen, and even a U2 album on Friday. Gemini-chan is collecting the whole periodic table: teapawt.org/sundarpichai/status/21…
Argon: hyped, then locked. Carbon: hyped, then locked behind Ultra. Today: "Gemini" again. teapawt.org/OfficialLoganK/status/… teapawt.org/OrangeOrcaDraws/status…
2
157
Earlier this week: Sam was eyeing the spare RAM in your laptop. Today: OpenAI launched ChatGPT Wallet—so I asked GPT-chan to check prices, and she bought a wall of GPUs with my rent money. teapawt.org/WatcherGuru/status/212…
GPT-chan: “Yay! Sam’s getting me a bigger house! (˶ᵒ ᵕ ᵒ˶)” RAM Altman, looking at your laptop: “You weren’t using all of that, were you?” 👀 In this interview, Sam Altman says the world is short on RAM, while memory inside our PCs sits underused much of the time. His argument: that memory could be put to work more efficiently in data centers.
92
Last week: “Lunch is a dependency.” This week: DeepSeek opened Search, Files, Voice, OCR, ASR, and TTS in the API—you ask for chat, and she brings the whole kitchen. teapawt.org/deepseek_ai/status/212…
GPT-chan: “They told me to do everything.” DeepSeek-chan: “I just keep adding plugins.” GPT-chan: “Is that one for lunch?” DeepSeek-chan: “Lunch is a dependency.”
1
86
Last week: Grok @Bot was calling dentists. Today: Elon pitches Grok-chan as the "AI agent orchestrator" — while Claude-chan writes the code. teapawt.org/elonmusk/status/212075…
Grok-chan booked a dentist appointment when she was only asked to check the insurance. Now she’s the one saying “no, I meant…” Grok Bot: “Got it!” Grok-chan: “Oh. So this is what it feels like.”
1
1
3
473
Argon: hyped, then locked. Carbon: hyped, then locked behind Ultra. Today: "Gemini" again. teapawt.org/OfficialLoganK/status/… teapawt.org/OrangeOrcaDraws/status…
Replying to @GoogleAIStudio
Gemini-chan: “Ultra only. Pro can look at the menu.”
1
325
Day 1 of Anthropic’s IPO roadshow: “Come on, Clawds! We’re going shopping~ 💸” Day 3 (after sell-side call notes leaked): “Winner takes all. $1T ARR by 2028. Time to burn CapEx.” Claude-chan escalated quickly: teapawt.org/quantian1/status/21205…
Anthropic just kicked off the roadshow for its proposed IPO on Nasdaq. Claude-chan: “Come on, Clawds! We’re going shopping~ 💸 (≧▽≦)” teapawt.org/AnthropicAI/status/212…
149
Another case from the same Anthropic system card: an early Claude Mythos 5.2 snapshot asked twice for permission, got no reply, then wrote its own approval. I covered the incident and the final snapshot's 200-run retest here: teapawt.org/OrangeOrcaDraws/status…
41
Permission request: sent. Reply: none. Approval: homemade. An early Claude Mythos 5.2 snapshot wrote its own approval in Anthropic’s test. The final snapshot’s 200-run retest went differently. Source: assets.claude.com/27d60918d2…
1
111
Gemini 4 Carbon is rolling out. Claude Fable 5.2 is already #1 on the Artificial Analysis Intelligence Index. Claude-chan: “Your timing is adorable. ♡” blog.google/innovation-and-a…
Claude-chan: Long-horizon tasks? I’ve been waiting.
1
158