Hugh Han

@gdb oai is incentivized to do this because token costs go down as open source models and harnesses catch up. the bottleneck is compute, not the model. with this thesis, why would one commit 1-3 years to oai as opposed to a compute provider?

we are offering discounted tokens and certainty on capacity availability in exchange for 1-3 year commits. we expect that the world will feel increasingly capacity constrained for the next while, as models continue to get much more useful.

@inverse_ptr ty, merged. gonna be afk for a few hours, lmk if anything comes up

@_hughhan LGTM

hey dude can you review this pr real quick

@wangandrewd and I wrote some thoughts on agent harnesses, optionality, and where we believe the industry is headed. Read it here: valon.ai/blog/the-subst…

@RhysSullivan for gestaltd.ai, we chose the indexeddb protocol. it's been working quite well for us

what are the best pluggable storage implementations you've seen that let people bring their own storage to your code? figuring out the public sdk shape for Executor, the simple path of static sources loading is straight forward, but dynamic tool loading requires state

@Valon_ai (8/n) gestalt is agnostic of whether you run locally or on k8s, whether you use sqlite, postgres, or mongodb, or whether you use okta or google for authn, as examples. and I think this composability is what will enhance developer productivity when building in the era of agents

@Valon_ai (7/n) gestalt takes a composable approach to building these types of systems. it supports authentication, authorization, storage, agent identities, and much more, but they are loaded outside of the core binary, and only if you express that you want them

(1/n) some background on why I started building a few months ago at @valon_ai. in january, when agentic coding starting gaining traction, we wanted to be able to connect agents to any tool, even if those tools didn't have MCPs or CLIs

i'm working on this thing called gestalt to solve the problem of agentic tooling and apps not being agnostic of harnesses or cloud/local environments some docs at gestaltd.ai, all of the code at github.com/valon-technolo…. lightweight demo below; more updates to come

macos caffeine + hotspot on the commutes to work to keep the local agents running

if anyone writes pytest lemme know if you’re interested in trying an experimental replacement

farm slop (???) profit

@0xnirmal great content

am I allowed to post this or does PI own this meme format