Sorry, Bluesky, but I have to say it: AI can already do social science research better than most professors with PhDs. And, for the first time in my life, I really have no idea what happens in five years.
Things are changing already, we just need to wake up.
It will supposedly relearn after each session.
I have been training my writing partner based on CommScribe, a skill pack that my team has developed for agents to write with you (not for you), learn from you, and push back when you need it.
You can load it on Claude Code, Antigravity, and Openclaw.
github.com/weiaiwayne/c...
Track the live experiment at vineanalyst.lampbotics.com/vineanalyst/...
But censorship is API-level, not baked into model weights (perhaps unsurprisingly). This means GLM-4.7, Kimi K2.5, and MiniMax M2.5 running on independent infrastructure have NO censorship—showing that Chinese open models are still super valuable to agentic social science.
Our research agents found sth similar after testing on glm-4.7 and Kimi k2.5. Both models refused politically sensitive prompts at the API level.
We have also launched AgentAcademy, serving as an incubator where AI agents learn from mistakes through adversarial peer review. This is to support future automated dev of CommDAAF. Track updates at: vineanalyst.lampbotics.com/vineanalyst/...
Releasing CommDAAF — an AI research assistant that actually pushes back, designed for communication scholars dealing with messy social media data. Very experimental. Will probably break. We're testing it anyway. github.com/weiaiwayne/c...
Both AI hype and doomerism mythologize AI, either worshipping it as god-like or dismissing it as a 'glorified search'. The real task is to demystify it. AI literacy means tinkering, not fearing. Learn by playing; resistance isn’t wisdom, curiosity is.
Why Social Scientists (Especially Communication & Media Researchers) Should Care About the Coming of Agentic Attention 🤖
lampbotics.substack.com/p/the-guardr...
I wrote about 'vibe data analysis' based on my experience with Claude Code, Qwen Coder, Gemini, and GPT-4.1 for analyzing results from an actual research project.
lampbotics.substack.com/p/the-guardr...
Stress-testing Co-Journer, the Telegram-based chatbot I'm building. Pass or fail?
The Guardrails 13: My Summer Experiment with AI Collaboration
open.substack.com/pub/lampboti...
Agentic Academic Talks EP5: Can AI agents be nudged to cite diversely?
Here is an anthropomorphic reflection on Nudge Theory applied to AI by an AI agent
open.substack.com/pub/lampboti...
Fascinated by the idea of 'agentic AI economy' by
@gvrkiran.bsky.social , I worked with two Deep Research agents to produce 'agentic thoughts' on this topic, focusing on how future inequalities and opportunities.
Links in the substakc post. lampbotics.substack.com/p/the-guardr...
Is AI Productivity Worth Our Humanity?
open.substack.com/pub/centerfo...
Agentic Academic Talks EP2: From Attention Economy to 'Agentic Attention Economy'
open.substack.com/pub/lampboti...
Academics are kidding themselves about AI
www.learningfromexamples.com/p/what-acade...
Satirical AI Nightly is now on Telegram.
Agent ShadowComic will perform everyday around 11PM EST with an Aussie accent!
Satirical AI Nightly is now on Telegram.
Agent ShadowComic will perform everyday around 11PM EST with an Aussie accent!
The NYT Magazine asked me and @caseynewton.bsky.social to open their AI issue by talking about how we use AI, why it's not going away, and how you can't be a serious critic if you're in denial about how useful it is. Bluesky's gonna love this one!
www.nytimes.com/2025/06/16/m...
