required pinned ggml patch & OMP,SIMD,etc. tweaks to make all models efficient, this work meshes well with other builds been doing with all-way sherpa onnx etc.
super cool roster! niche text generators come to mind, e.g., text version of regex train tracks / git traces come to mind. png/svg version https://regexper.com/#%5Cd%7B5%7D%28-%5Cd%7B4%7D%29%3F idk if any robust ascii lib exists, same for things like root cause or swimlanes etc.
have done some 1-off 1-shots of similar for AI input/output is primary frame of reference, and plenty of others have attempted same to varying success / nice to see ASCII graphics get some new life in them
interesting. tried it on iPhone SE. hard to do safari keeps thinking want to exit page. sliding one touch left right or u/d both might be good alt? certainly better than trad click exactly here on this little folder dart
one nice thing may be way to enhance penetrative ultrasound eg smaller immersive ultrasound tomography in smaller dip tank hey stick your leg in this thinner immersive tunnel , e.g., extremity up-close immersion or trunk wrappers vs just wand-based?
we can do tomography on any round-robin-rectify multi-pov source, doesn't have to be x-ray is just de facto use in medicine, closer to at min marketing ed problem
ya, brain just noisy channel in same way we can treat LLMs; anything possible exists we are just sampling it, which distills to "mere" clock syncing
L1 & L2 constraints unwind that clock compression with suitable dilation; very easy to think only efficiency matters and not averaged out replicas; nature does that inherently via primes, we have to create those artificial waves, recreate that convex hull, etc.
all to say, great to see more work in this direction & perhaps we can compare notes sometime!
one approach that can work is to tell model to load read skill and/or call shell script that overloads default, there are variety of ways to attempt this with any harness, claude specifically has hooks some of which allow go, no go, do this instead etc. and ya, agree on grokking code base, ast integration feels like natural next step
curious: wdym by "getting separators right when generating multiple files in a single inference call"
context: created hypertokens an even more robust hashing mechanism to create context-addressable memory (CAM), one cheat code is make them prefix-free, lots of others that get deep into why models work the way they do, etc.
we dug into those sorts of questions with hypertokens, a robust hash for lines, code, tables/rows or any in-context token tagging to give models photographic memory
one mechanism we establish is that each model has a fidelity window, i.e., r tokens of content for s tag tokens; each tag token adds extra GUID-like marker capacity via its embedding vector; since 1,2,3 digit numbers only one token in top models, a single hash token lacks enough capacity & separation in latent space
we also show hash should be properly prefix-free, or unique symbols perp digit, e.g., if using A-K & L-Z to hash then A,R is legal hash whereas M,C is not permitted hash
we can do all this & more rather precisely as we show in our arXiv paper on same; next update goes deeper into group theory, info theory, etc. on boosting model recall, reasoning, tool calls, etc. by way of robust hashing
Interesting! Been building space-time coordinate system for AI models. Notionally agree in principle w.r.t. convex hull, clocks, etc. since we invoke similar machinery albeit in tokenized models. Need read this work more deeply to grok.
One question is to what extent you dig into or have considered oversampling? One of the core hypotheses we've converged on is that nearly all models are optimized for source coding vs. channel coding. The implication is path to AGI likely involves oversampling to capture channel coding gains and which will resolve phase errors, etc.
Random sampling naturally does this albeit inefficiently. Curious if you do something more structured than random in terms of oversampling and especially partial overlapped samples / think supersaturated subspaces / subchannels, etc.
Basically, just thinking that it’s more ideal to have the tool call the micro VM versus the agent, doing it in the sense of its mandated by the tool call
interesting is the idea the agent calls it or just alt to terminal bash etc tool calls hey your tool calls are all microvms, containers, isoshells, raw term, clawd/molt all credentials with weaker and weaker security demarcs?
security matters if want to demarc where agents can play. running agent inside of strong VM is usually where starts container not enough for that full isolation only sees files you want it to etc
we've considered docker, firecracker, will add smol to working roster
context <> building something with QEMU
* required has to support LMW+AI (linux/mac/windows + android/ios)
there are scenarios in which we might spin micro vms inside that main vm, which by default is almost always Debian Linux distro with high probability.
one scenario is say ETL vm and AI vm isolated for various things
curious why building another microVM other than sheer joy of building, what smol does better or different, why use smol, etc. (microVMs to avoid etc also fair game :)
* https://sloop.ai/ * https://breezethat.com/ * https://drivespotter.com/
Socials: - x.com/SloopFX - x.com/DotDotJames - NYC, https://meet.hn/city/us-New-York
Interests: AI/ML, Science, Running, Space Tech, Startups, Technology, UI/UX Design, Physics
---
alum @riseofrest @techstars @tedx @usairforce