frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Show HN: Stream-based AI with neurological multi-gate (Na⁺/θ/NMDA)

https://github.com/CSCT-NAIL/CSCT
2•CSCT-NAIL•1h ago
Current LLMs struggle with compositional inference because they lack physical boundaries. CSCT implements a neurological multi-gate mechanism (Na⁺/θ/NMDA) to enforce L1 geometry and physical grounding. In my experiments (EX8/9), this architecture achieved 96.7% success in compositional inference within the convex hull—far outperforming unconstrained models.Key features:Stream-based: No batching or static context; it processes information as a continuous flow.Neurological Gating: Computational implementation of θ-γ coupling using Na⁺ and NMDA-inspired gates.Zero-shot Reasoning: Incurs no "hallucination" for in-hull compositions.Detailed technical write-up: [https://dev.to/csctnail/-a-new-ai-architecture-without-prior...]I’m eager to hear your thoughts on this "Projected Dynamical System" approach to cognition.

Comments

aghilmort•1h ago
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.

CSCT-NAIL•1h ago
Thank you for the profound insight. I completely agree that the path to AGI lies in channel coding (robustness and synchronization) rather than just source coding (compression).In CSCT, we don't just "sample" data; we process it as a continuous Projected Dynamical System. Here is how we address your points:

Structured Temporal Oversampling: Our stream-based approach effectively performs high-density oversampling in the time domain. Instead of random sampling, the theta-phase (hippocampal rhythm) in our MultiGate architecture creates structured, overlapping "integration windows" to capture temporal context.

Phase Error Resolution: Phase errors are resolved not by averaging (as in L2 models), but by NMDA-gating. The gate only opens when the anchor velocity and theta-phase align, physically "locking" the signal to a specific codebook vertex. This is a computational implementation of theta-gamma coupling.

Supersaturated Subspaces: Our Simplex constraint (L1) naturally handles what you call "supersaturated subspaces" by enforcing non-negative competition. This ensures that even with overlapping temporal samples, the resulting internal representation remains discrete and grounded within the convex hull.

By treating cognition as a communication channel between an "Anchor" and "Codebook," we prioritize the stability of the compositional mapping over the mere efficiency of representation.

China's Renewable Energy Revolution Is a Mess That Might Save the World

https://www.wired.com/story/china-renewable-energy-revolution/
1•xbmcuser•1m ago•0 comments

Show HN: ClawGate: Capability-based file access for isolated AI agents

https://clawgate.io
1•m64-64•1m ago•0 comments

Floppinux – An Embedded Linux on a Single Floppy, 2025 Edition

https://krzysztofjankowski.com/floppinux/floppinux-2025.html
2•GalaxySnail•3m ago•0 comments

Web Security: The Modern Browser Model

https://nemorize.com/roadmaps/web-security-the-modern-browser-model
1•reverseblade2•3m ago•0 comments

The Impatient Programmer's Guide to Bevy and Rust: Ch. 7 – Let There Be Enemies

https://aibodh.com/posts/bevy-rust-game-development-chapter-7/
1•kelvie•8m ago•0 comments

Coding assistants are solving the wrong problem

https://www.bicameral-ai.com/blog/introducing-bicameral
1•jinhkuan•11m ago•0 comments

Ask HN: Request limits vs. token limits for AI-powered apps?

1•JeduDev•12m ago•0 comments

Ask HN: Tell me about your internal tools

2•doawoo•14m ago•0 comments

Prompt and Context Engineering App – My First SaaS Launch – ImPromptr

https://impromptr.com
1•saddatahmad•15m ago•1 comments

Planning-with-files: Claude Code skill implementing Manus-style workflow

https://github.com/OthmanAdi/planning-with-files
1•rahimnathwani•16m ago•0 comments

Ask HN: Is anyone losing sleep over retry storms or partial API outages?

1•rjpruitt16•18m ago•0 comments

Use Less Shit

https://uselessshit.org/
1•cvalka•19m ago•0 comments

A Pragmatic VLA Foundation Model

https://arxiv.org/abs/2601.18692
1•mountainview•19m ago•0 comments

Taiwan's orange travel alert for Hong Kong, Macau remains for Lunar New Year

https://www.taiwannews.com.tw/news/6294341
1•maxloh•22m ago•0 comments

Longest non-life prison sentence ever handed down to a single person?

https://ballparkguess.com/?id=60280c22-7c29-4b36-a1fb-ff4a6d052ebc
1•bkls•23m ago•0 comments

Frog 'saunas' could help endangered species beat a deadly fungus (2024)

https://www.science.org/content/article/frog-saunas-could-help-endangered-species-beat-deadly-fungus
1•noleary•24m ago•0 comments

Show HN: Kannada Nudi Editor Web Version

https://nudiweb.com/
2•Codegres•25m ago•0 comments

Innovation Biopics

https://arbesman.substack.com/p/innovation-biopics
1•surprisetalk•28m ago•0 comments

Viewing Forth

https://programmingsimplicity.substack.com/p/viewing-forth
1•surprisetalk•28m ago•0 comments

33rpm – A vinyl screensaver for your Mac

https://33rpm.noonpacific.com/
1•surprisetalk•28m ago•0 comments

The Recent 0-Days in Node.js and React Were Found by an AI

https://winfunc.com/blog/recent-0-days-in-nodejs-and-react-were-found-by-an-ai
1•mufeedvh•32m ago•0 comments

The Coherence Premium

https://www.joanwestenberg.com/the-coherence-premium/
1•zdw•37m ago•0 comments

Zai/GLM-OCR

https://huggingface.co/zai-org/GLM-OCR
2•naze•41m ago•0 comments

xAI engineer fired for leaking secret "Human Emulator" project – Macrohard

https://www.youtube.com/watch?v=0hDMSS1p-UY
4•consumer451•47m ago•0 comments

Predicting Math.random() in Firefox Using Z3 SMT-Solver

https://yurichev.com/blog/xorshift/
2•thunderbong•52m ago•0 comments

AI After Drug Development

https://asteriskmag.substack.com/p/ai-after-drug-development
1•crescit_eundo•53m ago•0 comments

Opus 4.5 designed a political Magic the Gathering set

https://github.com/FeSens/Mirrodin-Manifest
1•fesens•54m ago•1 comments

Why AI Swarms Cannot Build Architecture

https://jsulmont.github.io/swarms-ai/
2•gmays•54m ago•1 comments

Termix: Web-based platform with SSH terminal, tunneling, file editing

https://github.com/Termix-SSH/Termix
1•thunderbong•58m ago•0 comments

Ask HN: Where do all the web devs talk?

2•LinguaBrowse•59m ago•0 comments