3. Agent Demo
A Chinese chat about CP, entertainment gossip, Stephen Chow's Kung Fu, and a smile imitation.
Voice: cheerful female with laughter
Scene: bright white indoor room
Tried Pioneer before as inference provider, was not happy with reliability and quality of output. another story with umans! I can have multiple concurrent sessions (max 4) at decent TPS (~50) for $50/month flat rate
I believe what Anthropic is doing, gating the ability to do certain harmless things like LLM research, and with incredibly sensitive filters that even medical questions are often blocked, is *deeply* wrong. They got open research, the Transformer, GPT2, ...
@zeeg I’m working on something similar for pi coding agent tools, in particular for bash tool calls it parses the state at each point of the tool calls via the `unbash` parser
Update: it works! Just drop the rule in ~/.pi/steering/index.ts and hot /reload
The steering hook analyses bash commands with unbash, tracks working dir, env vars inside each tool call, then if the rule matches it blocks the command and replaces tool output with the rule
Recommended reading from @mitchellh
https://t.co/yXfdM0C4mM
High quality open source primitives that are easily slop-extensible for user personalisation