I ran a comparison of @ornith_ model 1.5 MoE in @goose_oss and in @MiniMax_AI new CLI harness.
Damn near close. @MiniMax_AI built a fine harness. But that also means @goose_oss deserves more attention than it receives.
@singularity_sah@DanielLockyer Does GLiNER answer the same three types of questions Jev does? I asked Gemini about this and it says the use cases for GLiNER and Jev are different.
@JD_DeYonker@ItsMikeBloom Single family residential and small commercial is largely prescriptive and deterministic tools already solve for most the client needs. Larger commercial real estate is more complicated and less templated, which means more reasoning and not solved by software vendors.
@HermesWatcher Went from maxxing Telegram bots to one bot across multiple topics. Used a single Hermes agent with LLM that would message me, and when I replied, would route to the best-fit LLM's and persona's.
Slack + Claude Tag is really good, but it’s very token hungry and ofc is limited to just Anthropic models
Buzz + [your model of choice] just makes so much sense, on both an economical and freedom basis
Provision tools and MCP's by channel rather than by agent in Buzz (forked). Agents work across channels, retaining knowledge and lazily load what they need on-demand. Build cross-domain expertise with more skills than they can ever access pre-loaded.
@mr_r0b0t@theami_ Oh this would have saved me HOURS on Friday, porting my Hermes agents over to Goose. Bravo!
I am running a fork of Buzz with MCP's and skills point and click assignable by channel rather than by agent.
FOSS is amazing.
Lots of talk of “moats” in my X feed. I predict citadels and castles to trend once the dust settles for the few winners. Could be months. Could be years. Could be weeks from now at the rate AI is aging. Who knows!