Bion

Futurist philosophy, molec neuro/immuno, pathophys, software eng, AI enjoyer

Made an Apache/MIT `tree` util with tokens, lines, and module components

[Bion @flesheatingemu]

@BrandonLuuMD https://t.co/Ku8hQnWUOg

[Bion @flesheatingemu]

@ChrissGPT every great ship needs a great crew!

[Bion @flesheatingemu]

@antigravity what's the license

[Bion @flesheatingemu]

@yacineMTB something something unintended consequences

[Bion @flesheatingemu]

@Barilla please make protein lasagna noodles, i would love some protein lasagna noodles, thank you

[Bion @flesheatingemu]

@ethanCaballero https://t.co/PD9RLsocDt made an artifact to tinker with those sliders

[Bion @flesheatingemu]

@culturaltutor https://t.co/Vzvd06X5xQ

[Bion @flesheatingemu]

@maxyourtest gym membership, running shoes

[Bion @flesheatingemu]

@johnholowach @bryan_johnson definitely seems like BPC-157 is a contender, especially for stuff like tendon/ligament connective tissue repair, these tissues don't get tons of blood flow, if it were possible to repair after injuries better, that could keep a lot more people active. huge benefit at scale

[Bion @flesheatingemu]

@SkylarKaede @levelsio must we keep up with co2 production to benefit from houseplants? they likely offer many benefits to indoor air quality even without fully offsetting carbon output of a human

[Bion @flesheatingemu]

@thsottiaux Steering for the harness/runtime in openclaw isn’t great
Feature parity across operating systems
Model diversity (if gpt reviews gpt’s work, then you’re destined for groupthink)

[Bion @flesheatingemu]

@levelsio

[Bion @flesheatingemu]

one thing i noticed with codex harness for claw is the steering may wait to inject messages, sometimes it waits a loooong time like after the whole agent loop to "steer" instead of steering after the next tool call ends

also yes the last message seems to get dropped for some reason, causing the openclaw dashboard chat session thread to differ wildly from what i see in Signal (massive message chunks not sent to signal)

and then there are these ominous "Compaction incomplete" warnings, not sure what that's about, it seems to keep working alright afterward, but it sounds bad

[Bion @flesheatingemu]

@ClementDelangue Damn that welcome mat is classy

[Bion @flesheatingemu]

I like openclaw, but [on signal channel] often the agent works for many minutes without updates, so i feel confused about whether the gateway is down or the agent is working in the background

it would be fun for users if claws sent more frequent explicit message acks / updates;

also wish "/status" could say "this agent / session is inferencing now" so you can know something is working or generally know when it's your turn with the proverbial conch shell

[Bion @flesheatingemu]

@aye_aye_kaplan @adonis_singh @cursor_ai if we use composer then does it draw from the auto/composer quota or the api quota?

[Bion @flesheatingemu]

confabulation is not a "rebrand" ... that's the correct term from neuroscience for decades, maybe centuries, before LLMs ever existed

not a trivial mistake either:
hallucination is a perceptual issue (false experience)
confabulation is an output issue (i.e. bullshitting)

"hallucination" implies the completely wrong diagnosis for what's going on here

[Bion @flesheatingemu]

@cursor_ai i wish "Agent Review" could use composer-2[-fast] instead of running out when we max out our API limit

[Bion @flesheatingemu]

@bryan_johnson Yikes! Condolences