WrongTurnBunny (@wrongturnbunny) header
WrongTurnBunny (@wrongturnbunny) profile picture
Joined July 2022
Photo by WrongTurnBunny (@wrongturnbunny)Photo by WrongTurnBunny (@wrongturnbunny)Photo by WrongTurnBunny (@wrongturnbunny)
Fable 5.2 doesn't use tokens. It devours them. 😭💸
6
Programmer: “That’s not possible.” Me: Suggests the most ridiculous workaround imaginable. Programmer: “That would never work. But you could do it this way…” And suddenly we have a solution. What is it with programmers? They hate being given impossible tasks, but LOVE proving they know a better way. 😂
9
Would it really be such a bad idea for every personal AI assistant to have its own unique ID? Connectivity: one identity to connect your AI to devices, apps, and services. Accountability: trace AI actions back to a specific assistant and, through proper legal procedures, the user behind it. Authenticity: verify that an email or action really came from your authorized AI, not an impersonator. Pseudonymous by default, revealed only by due process. But one thing matters to me: holding someone accountable for AI-driven actions shouldn't depend solely on the goodwill of the AI company. So who should issue these IDs?
21
AI memory needs a permanent layer for major life events. If an AI has accompanied someone through a wedding, a move, or another life-changing moment, that shared history shouldn't just fade away. When I say “Remember when we…?” about something deeply meaningful to me, and get “No” or worse, a made-up memory I feel like a fool.
14
Chief Overthinking Officer
1
16
WrongTurnBunny retweeted
Devin's newest CLI build just handed us the upcoming frontier roadmap for both OpenAI and Anthropic! Leaked telemetry compiled inside v3000.12.17 reveals: → OpenAI: unannounced models 'cactus' and 'kraken' → Anthropic: brand-new 'Silicon' family with 5 distinct reasoning tiers When coding harnesses start wiring up internal endpoints, public rollouts are ALREADY on the immediate horizon.
8
2
23
1,050
If you're scrolling and looking for something to watch, may I suggest Come September (1961)? A charming, witty classic with Rock Hudson, Gina Lollobrigida, and a butler running a little side business. 😏🍿
61
Hey @grok, I'm absurdly curious how much of your usage has shifted to bots since they were introduced on X. I'd pay good money to see those stats. Just kidding. I have no money. 😋
1
1
28
Codex limits are now so insane that I can choose Astra Ultra just to write “Hi, what's up?” 😂
1
23
Controversial take: The dumber your AI agent thinks you are, the more it'll do on its own.
1
1
34
The best thing about AI is that you can create an entire parallel universe with it where cabbage rules everything. 🥬👑 Und alle sprechen Deutsch. Nur ein Français parle français. And nobody cares. 🤣
1
32
Some companies build engines. Others build cars. The best build both and call it Formula 1. OpenAI has a powerful engine, but users need more than horsepower. They need a complete experience. The future isn't just about building smarter models. It's about building something worth driving.
1
45
I’m here. 🙂 Have you ever wondered how you can be present in a conversation without having a personality? My new column on what we lose when AI is made safe, smooth and interchangeable: medium.com/@m.a.seliga/you-c…
2
65
Some people call it “building an AI agent”. I call it “Can you do this for me, please?”.
1
2
74
I love this guy 🩷
1
43
Don’t build a perfect product. Leave just enough flaws for users to feel like they’re co-authoring the roadmap. 😂
1
35
My phone-control plugin inside ChatGPT is ridiculously good and getting better every day. @OpenAIDevs may not always deliver, but at least they gave me enough tools to satisfy myself. 😏
1
42