St'Unquo

A human account for rough draft rethinking in the AI augmented future of everything.

[St'Unquo @status_unquo]

@froehlichmmm @karpathy I understand openclaw's marketing cracked through to the masses who aren't keeping tabs on the opensource landscape (many discovering github for the first time / never browsing other repos) but they weren't unique

[Anthropic @AnthropicAI]

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.

The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.

Access to all other Claude models is not affected.

We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.

Read our full statement: https://t.co/bwn0sximKZ

[St'Unquo @status_unquo]

@codevsdev Ok so hear me out: by coding.

[St'Unquo @status_unquo]

@araseb_ would an AI typo "strobe"? (I mean... yes, probably to faux human... but anyway..)

[St'Unquo @status_unquo]

@araseb_ on a video call - you can stobe light into your camera in random colour sequences at random rhythms, and my face will reflect your patterns without any change to latency

[St'Unquo @status_unquo]

Good morning Claude quota reset! Time to test whether Fable 5 can eat an entire week before breakfast tomorrow?

[St'Unquo @status_unquo]

@Jeremy069626792 So many people starting to share your project! But took me a hunt to find you on X. Do you have a website / anything else I should be following ontop of Github?

[NuClide @n15647931]

BARE
one ~101 MB Rust binary.
Embedded BERT.
Zero Python. Zero deps. Zero network.
Drop it anywhere (air-gapped, SCIFs, classified) and get semantic Metasploit module ranking from:
• Nmap
• Nuclei
• Shodan
https://t.co/3liCCJerIj

[St'Unquo @status_unquo]

Finally tracked down jcode's OG developer @Jeremy069626792 ! One hour in - feels like the performant + resource efficient coding harness we've yearned for. Early days, keen to see this gain traction... make sure to follow him here + login to star the Github repo (currently 2.3k) because this deserves to rise through the slop star farmer noise.

[Om Patel @om_patel5]

SOMEONE JUST DROPPED A NEW AGENT HARNESS THAT CLAIMS TO BLOW PAST CLAUDE CODE AND CODEX

quick context: an agent harness is the layer between you and the AI and it handles prompts, tools, memory, and parallel sessions (e.g. codex cli & claude code)

it's called jcode AND it's free and open source on github

claims to be:
> 20x more memory efficient than claude code
> spawns 63x faster than codex CLI
> more customizable than every existing harness
> built in memory system, no extra setup
> built specifically for running parallel sessions and agent swarms
> handles background tasks, browser use, and codebase searching natively

the part everyone is losing their minds over is that you can log in with your existing claude code or codex oauth and just use it as the harness

so you keep your max plan, but you run it through a faster, more efficient interface that can spawn 20 agents at once if you want

the catch is that anthropic has banned people for using claude oauth with third party tools. so this lives in a grey zone right now

but if you're already running multiple claude code sessions in tmux or paying for codex separately, this is built to replace both with one harness

the open source community is moving way faster than the big AI companies. again.

[St'Unquo @status_unquo]

If enabling distributed/decentralised training to scale to frontier models is the goal. How might billions in funding change how we get there? Make device/OS/hardware ubiquitous that bakes in sharing latent/idle compute, normalise participation among average consumers? Solve for the economics of increasing access to prosumer compute, for more people to participate? Make decentralised / open source solutions intuitive enough, without technical aptitude, to become a consumer norm? Wield the power to disrupt current power structures, to recentre on new mental models that don't centre technologies around commercial "apple moments".

[Kat ⊷ the Poet Engineer @poetengineer__]

trying to use topological data analysis to map the shape of my x bookmarks through mapper + embedding extraction and generated 3 views:
- density: where attention keeps gravitating
- pca: the dominant axes of variation
- centroid: center vs edge (typical -> outlier)

[Odyssey @odysseyml]

It’s time to go beyond language models.

Introducing Odyssey-2 Max, our most powerful world model yet. It materially advances the SOTA in physical accuracy.

This is a big step toward models that simulate and interact with the world in real time.

A new intelligence entirely!

[St'Unquo @status_unquo]

LLMs might be argued to process the mechanics of some part of emotion, however their substrate does not feel it. Airplanes perform the mechanics of flying, but their substrate does not feel it.

Even in humans - there's a strong school of thought our emotions are not our thoughts? and may not arise solely from the state of the brain?

Emotion feels like a whole body experience of persistant physiological states changing over time, responding to stimulus that may or may not include concious thought.

Not just the mind generating the string "I feel sad" or "this is amazing", which might even contradict the felt emotion.

With most training data still constrained to what we have given language to, it doesn't seem LLMs are there (yet?).

[Zain Shah @zan2434]

Imagine every pixel on your screen, streamed live directly from a model. No HTML, no layout engine, no code. Just exactly what you want to see.

@eddiejiao_obj, @drewocarr and I built a prototype to see how this could actually work, and set out to make it real. We're calling it Flipbook. (1/5)

[St'Unquo @status_unquo]

@fat The "boring" things are the most exciting thing teams can build in 2026.

[St'Unquo @status_unquo]

@PSYCHOV3N0M @VictorTaelin Off the top of my head you could join the Discords for: Nous Research (DisTrO), Prime Intellect, Petals. Bittensor, Gensyn, io net. Pluralis Research are solving for model parallelism vs data parallelism, for pretraining, tho early days and not sure when their next run will be.

[St'Unquo @status_unquo]

If it can be built in a weekend today, it will be spun up by a prompt tomorrow.

So.. commercial projects might need to design for a reality where every user can effectively-enough solve most of their needs.. but pay a premium for solution design they trust to do the thinking for them, and enable a communable experience if also trusted by their peers.

Be that "trust" in security/reliability, or "trust" to stay ahead of the curve.

[St'Unquo @status_unquo]

Still baking my view on this. Maybe we need tight architectures, spec driven development. Modularised code where speed:quality can be granularly prioritised.

Maintain structural rigour with some tolerance for "mess" within guardrails - where short feature half-life demands velocity of change, and/or risk is low.

Elephant in the room being whether today's notion of full-stack ready-made "apps" has its own half-life. Maybe we need to be designing toward playing nice with "messy" user generated code and/or code spun up ephemerally with no mind to maintenance.

Needing to use a fixed separate "perfect" ready-made UI/UX, might become more frustrating for users than "good enough despite some minor bumps" integration with their own interface/surface/harness/whatever-we-are-calling-it.

Which is also to say anything devs are building in a few hours today, users will soon just spin up as a prompt themselves. So agree engineering effort needs to focus on design rigour rather than just what is easily shipped.

[St'Unquo @status_unquo]

@JinjingLiang @orca_build because we want open standards and interoperability. to use any agent/CLI even if the interfacing product hasn't shipped native integration, to triage and change between them. to build flexible consumer expectations early, to push back against normalising vertical stack lockin.