Practical Physicist aka Super Genius, Dad, Jack-of-all-trades, master of two. CEO/Founder of @gnusai
California
- Tweets 1,559
- Following 2,123
- Followers 2,145
- Likes 219
Pinned Tweet
Ok, here it is. This is game-changing, and what we are about to build and release on top of the @gnusai Operating System.
drive.google.com/file/d/1iuJ…
Ambitious? Maybe... But with our OS already built, IMHO, a 3-6 week project. With embedded proven blockchain technology.
Humanity, you are welcome.
@elonmusk, @PeterDiamandis "A day before something is a breakthrough, It's a crazy idea", @JeffPeoples, @beffjezos, @XFreeze, @tetsuoai, @BetterCallMedhi, @EricJorgenson, @gailalfaratx, @cboyack, @MarioNawfal, @Natortelli, @SawyerMerritt,@michaelnicollsx, @MrBeast, @Cobratate, @SawyerMerritt, @grok, @rohanpaul_ai, @Chad_Hurley, @stevechen, @sama, @brycent, @Ronald_vanLoon
So true, why we are building something unique.
The GNUS Cognitive App is a system that continuously models, challenges, and improves an individual’s thinking.
Built on top of @GnusAi a distributed, verifiable compute network that executes and coordinates arbitrary workloads—including AI, rendering, and cognitive processes—while continuously optimizing cost, performance, and correctness.
7
7
29
4,911
Kenneth Hurley retweeted
C'mon man...
The 2013 AMOLF image maps the probability density of an electron wave function in a highly excited hydrogen atom using photoionization microscopy. The 2015 EPFL image maps light (plasmons) confined on a nanowire by watching how electrons exchange energy with those photons.
Both produce standing-wave or interference patterns because both electrons and light behave as waves when confined. Similar-looking rings or intensity maps appear in many wave systems; the resemblance does not mean one object is made of the other.
Standard physics already treats electrons as having wave-particle duality and rest mass, while photons have neither rest mass nor charge. Visual similarity of two laboratory reconstructions is not evidence that matter is confined electromagnetic radiation.
1
1
216
Why is AI-assisted code so valuable? Here's why, and this is not Vibe coding.
Working on a newfangled algorithm. Probably only a handful of people would understand what this bug means, as it requires a deep understanding of GPU hardware, shading, and rendering. It's probably Gobbleygook to most people.
"The claim is confirmed by code reading: CSReduce divides by valid-child count at every level, so when a parent mixes full and partial blocks (any level where children have unequal texel footprints — e.g. 129→17→3 ends with a 3×3 combine of blocks covering 4096, 64, and 1 texels each weighted 1/9), the result is a block-mean, not a texel percentile. Level 1 is exact (per-texel count loop); the corruption is levels ≥ 2. POT sources never mix (all weights equal — which is why the existing tests pass).
There's an exact fix that stays surgical: give the kernel each child's raw-texel footprint weight, normalized by span² so stored values stay ≤1 and nothing overflows the half-float ping-pong targets. Level 1 degenerates to weight 1 — current behavior. Checking the plumbing and test harness before writing it:"
I've had to guide AI in diagnosing bugs based on my years of experience and intuition.
1
1
14
272
@grok explain for a layperson to understand what the bug means.
Another good use for AI.
1
4
71
Kenneth Hurley retweeted
Granularity matters. AI tokens, yes; NFTs, yes; maybe even RWAs like electricity. But AI encapsulates a lot, and things like electricity, inference, etc. get aggregated as a component of the token. @GnusAi
3
11
142
BlackRock paper describes agentic AI using MCP/A2A for coordination, x402/stablecoins for M2M payments, and agents autonomously discovering/provisioning compute via marketplaces (Figs 2/5), with blockchain settlement and metering.
GCS/GNUS already implements specialist ELMs, swarm orchestration via P2P/CRDT job queues, reputation-weighted routing/discovery, distributed inference on edge nodes, EIS verification, and hybrid blockchain escrow/settlement in GNUS tokens.
Direct matches: specialists, compute sourcing/routing, metering, verification, reputation, autonomous settlement.
GCS exceeds in cognitive memory/grounding, local-first privacy, and edge focus. Gaps (x402/A2A/MCP, KYA, tokenized claims) are protocol adapters, not core redesigns.
Core architectures are very close; BlackRock outlines a market vision GCS was built to enable.
1
10
133
Kenneth Hurley retweeted
@grok Analyze this BlackRock paper, “The Machine-Native Economy,” and compare its proposed architecture with Genius Cognitive System (GCS) and GNUS.AI using:
* gcs.gnus.ai/llms.txt
* docs.gnus.ai/llms.txt
* github.com/GeniusVentures/Ge…
Focus on the paper’s ideas around autonomous agents, specialist agents, compute discovery and routing, distributed inference, machine-to-machine payments, blockchain settlement, metering, verification, reputation, MCP/A2A/x402-style protocols, and agents autonomously sourcing and paying for compute.
In particular, examine Figures 2 and 5 and map each major component to any equivalent already present or planned in GCS/GNUS.AI.
Do not assume they match. Determine from the architecture and documentation whether the BlackRock model is substantially different, partly overlapping, or whether the core architecture BlackRock describes is very close to what GCS + GNUS.AI were already built to do.
Identify:
1. direct architectural matches,
2. areas where GCS/GNUS goes beyond the BlackRock model,
3. genuine gaps such as x402, stablecoin settlement, agent identity/KYA, A2A, or tokenized compute claims,
4. whether those gaps require architectural changes or are mainly adapters/protocol integrations,
5. and your final technical conclusion on how close the two architectures really are.
Be skeptical and evidence-based. Quote or cite the relevant parts of the BlackRock paper and the GCS/GNUS documentation rather than accepting the comparison at face value.
5
10
24
400
Kenneth Hurley retweeted
@Grok Analyze arxiv.org/pdf/2506.15121 and compare thermodynamic/Langevin computing with photonic computing.
Would using optical noise, interference and coupled optical systems give the same stochastic-computing benefits with better speed and energy efficiency than thermal hardware?
Include relevant recent research.
2
1
7
406
Working with AI agents has let me not only work on AI but also prove the concept of some cool ideas for graphics techniques à la ShaderX and GPU Zen @wolfgangengel. @GnusAi- I'm still 99% working on AI and the apps, though.
raw.githubusercontent.com/Gr…
2
4
15
635
🚀 Excited to be speaking at the Global Innovation Showcase 2026!
I’m looking forward to joining founders, investors, government leaders, technology pioneers, and innovators in San Francisco for a day of conversations around the technologies and ideas shaping what’s next.
From AI and emerging technologies to Defense Tech, Smart Cities, GovTech, infrastructure, and cross-border innovation, the Global Innovation Showcase will bring together perspectives from across industries and global ecosystems.
📅 October 6, 2026 | 11:00 AM – 7:00 PM
📍 San Francisco War Memorial
🎟️ Register here: partiful.com/e/9WxqlWmhozblw…
🌐 Hosts: The Future of Tech (Orbis86, OffChain Global, BrandPR) | Silicon Valley PropTech Association
🤝 Beverage Partner: Gnus.ai
Looking forward to being part of the conversation and connecting with everyone in San Francisco!
7
16
32
834
While I don't agree 100% with everything in this article—like not reviewing code; IMHO, being the "human in the loop" is still very necessary—this part is what I've been telling my engineers:
"If all you do for your paycheck is fit into a process where other people decide what to do and you transform those directions into software, congrats, you're a 'meat proxy.' If all you do is take direction from higher-level bosses and transform those into instructions for other people, you're also a 'meat proxy.'"
Your job may be surviving on inertia. Your resentment won’t extend the runway. What do you actually want to do with the intelligence now at your disposal?
1
7
366
Kenneth Hurley retweeted
Imagine that I and Steve Coallier had to force that 29mb game into 3.5mb on the Playstation 1!
2
5
82
Kenneth Hurley retweeted
AKA distributed compute and we're getting ready to launch exactly that. gcs.gnus.ai
1
6
13
143
Exactly what @GnusAi is the "Globally Networked Ultra-Smart Artificial Intelligence" is a full AI Operating System platform
1
6
17
470
Glad to see "Human in the Loop" is very much needed for coding.
-- That's a real production defect in the D-04 reprocess path — your instinct found it.
1
1
8
225
My charter too. @GnusAi swarm data center at scale
Literally right this minute, my mini data center is running over 20,000 tokens / second (aggregate), processing science papers, articles, transcripts, PDF files, scans and more for use in my curated document index that powers BrightAnswers.ai (a free AI research engine that offers accurate citations from real research documents).
NONE of this would be possible without open source AI. In fact, 100% of these tokens right now are being processed by Qwen3.8-27B including VL (vision language) processing of page scans.
I spent over $2 million putting all this together over the last 2.5 years, but today this entire effort could be done at 1/10th that cost thanks to open source AI.
My engine is free for you to use, even though it costs my company money to host it, because I believe in paying it forward. I use open source software, and I build free-to-use platforms that benefit humanity.
Lets all do MORE of this!
4
14
298