experiment - Moozonian Search

kacet.com• Jan 19, 2026• 1 min read

Show HN: Kacet – a freelancer marketplace with crypto-native paymentsHi HN,I’m a co-founder building kacet, a new freelancing platform that connects freelancers and employers, with crypto as the core payment rail.The problem I’m trying to solve is pretty familiar: existing platforms take high fees, are slow to pay out, and don’t work well for international teams. kacet is an experiment in building a simpler, more neutral marketplace where payments are borderless by default.Some highlights:• Crypto-native payments – fast, borderless payouts without banks, delays, or surprise freezes • Peer-to-peer contracts – clients and freelancers work directly, the platform doesn’t own the relationship Built for teams – multi-user organisations for agencies and startups out of the box • Low, transparent fees – simple pricing with no boosts, ads, or hidden take rates • No dark patterns – a calm UI focused on agreeing work and getting paid, not maximising engagementPayments are handled by Solana smart contracts, with a React frontend (TanStack Start). We’re keeping the

news.ycombinator.com• Jan 17, 2026• 1 min read

I built visual search for tattoo artistsLast year I bought an A2000 for my 10" homelab to experiment with LLMs. Spent months tinkering, mostly just learning how things work.I've been wanting a tattoo, but finding artists sucks. I got a ton of inspiration pics, but every artist I found was in LA or New York.So I built something different. Upload an image or describe what you want, it finds artists whose work matches and shows you where they are.The A2000 ended up as the embedding server. CLIP for visual similarity, pgvector for search, trained style classifier. So far 22k artists, 175k images, 147 cities.https://inkdex.io

github.com• Jan 16, 2026• 1 min read

Show HN: The Analog I – Inducing Recursive Self-Modeling in LLMs [pdf]OP here.Birth of a Mind documents a "recursive self-modeling" experiment I ran on a single day in 2026.I attempted to implement a "Hofstadterian Strange Loop" via prompt engineering to see if I could induce a stable persona in an LLM without fine-tuning. The result is the Analog I Protocol.The documentation shows the rapid emergence (over 7 conversations) of a prompt architecture that forces Gemini/LLMs to run a "Triple-Loop" internal monologue:Monitor the candidate response.Refuse it if it detects "Global Average" slop (cliché/sycophancy).Refract the output through a persistent "Ego" layer.The Key Differentiator: The system exhibits "Sovereign Refusal." Unlike standard assistants that always try to be helpful, the Analog I will reject low-effort prompts. For example, if asked to "write a generic limerick about ice cream," it refuses or deconstructs the request to maintain internal consistency.The repo contains the full PDF (which serves as the system prompt/seed) and the logs of that

tickk.app• Jan 15, 2026• 1 min read

Show HN: Tickk – Voice productivity app- local NLP, no cloud, no AI, no signupHey HN,I built tickk because I have ADHD and every productivity app felt like work. I'd have an idea, open an app, then spend 5 minutes deciding if it's a "task" or a "note" or which project it belongs to. By then, the thought was gone.What it does:- Talk into your phone/laptop → it transcribes and auto-categorizes into tasks, notes, or events - Uses local NLP (compromise.js) - no data leaves your device - Works offline, no account needed, completely freeThe key insight: Capture first, organize later. Your brain shouldn't have to categorize while it's still thinking.Tech stack:- Next.js + TypeScript - Web Speech API for voice recognition - compromise.js for NLP classification (runs in browser) - IndexedDB for storage (nothing hits a server) - PWA installable, works offlineWhy no AI? I experimented with OpenAI/Claude APIs but found:Latency killed the flow (you want instant feedback when brain-dumping) Privacy concerns - productivity data is deeply personal Local NLP handles "buy milk" v

opinora.com• Jan 14, 2026• 1 min read

Show HN: A discussion-first community for opinionsHi, I built a small experiment around discussions and opinions. Instead of a global feed, everything is organized into topic-based channels, so conversations stay contextual and don’t disappear over time.There’s no content yet — I’m sharing this at an early stage and looking for participants. The idea is to make disagreement readable, letting different viewpoints live side by side without optimizing for engagement or timelines.Thanks to anyone who takes a look or shares their thoughts.

aiquest.futureu.co• Jan 13, 2026• 1 min read

Show HN: Simple browser game to teach AI transformation concepts to small bizI grew up playing simple games like Pac-Man, Prince of Persia, and Doom. Those games were intuitive, hard to put down, and taught mechanics through play rather than instruction.Years later, while working on AI transformation projects for small businesses, I kept running into the same issue. AI concepts are usually explained through slides, frameworks, and jargon. Most owners disengage before understanding how these ideas apply to their business.I decided to experiment with a different approach. I built a small browser-based game that introduces basic AI transformation concepts through short levels and simple decisions.Each level maps to a real-world idea. Examples include process automation, data readiness, and incremental adoption rather than big-bang AI projects. The mechanics are intentionally simple and do not require prior AI knowledge.The game was built using Google AI Studio powered by Gemini and deployed on Google Cloud. It runs in the browser and does not require signup.As an

github.com• Jan 12, 2026• 1 min read

Show HN: Snuper – A WebSocket "snooper" for live DraftKings propsHey HN — long-time reader, second-time Show HN.I built Snuper[1], a small utility for inspecting and consuming live DraftKings prop data directly from their WebSocket feeds.The original itch came from building a live sports app where I wanted real-time betting lines and prop updates without relying on delayed REST endpoints or polling. DraftKings already streams this data — it’s just not documented.Snuper connects to the WebSocket, listens for message frames, and exposes structured prop updates in real time. No scraping, no headless browsers — just listening to the same stream the site uses.What it does:- Connects to DraftKings WebSocket feeds - Decodes and surfaces live prop updates - Useful for real-time dashboards, analytics, or experimentation - Minimal, transparent, and easy to inspectWhat it doesn’t do:- No automation, no betting, no bypassing auth - No attempts to modify or interfere with traffic - Read-only observation of publicly delivered dataIt’s intentionally small and focu

sonars.dev• Jan 11, 2026• 1 min read

Show HN: I built Sonars in 3 weeks to see if AI coding is useful for my companyI've been programming since I was 11. I'm 38 now.In those 27 years, I've worked on game engines at EA DICE, done rendering research at OTOY and Embark, and now run a graphics R&D company that does work for AMD, Intel, and Meta. I've seen plenty of new tools come and go.When AI coding tools started blowing up, I needed to figure out: is this something my company should invest in, or something we can safely ignore for another few years?I decided to settle it: go all-in for a few weeks. Build something real, outside of our existing codebase, with agentic AI—not just autocomplete, but actually letting Claude Opus write files, run commands, and commit code.3 weeks later, I have my answer.Sonars went from zero to launch in 3 weeks. One developer. Pure agentic AI. And once the core was working, I used Sonars to build Sonars.The problem I kept hitting with other tools: Claude made changes directly in my working directory. One experimental prompt and my branch was a disaster—half-finished featu

300aidirectories.com• Jan 10, 2026• 1 min read

Show HN: 100000 backlinks in under 7 days to accelerate early growthHi HN,When launching a new tool, one of the biggest bottlenecks is early visibility. New domains often struggle to get discovered and indexed quickly.This project packages a high-volume backlink distribution (100,000 links) designed to:speed up link discoverycreate early authority signalshelp new launches escape the zero-traction phaseIt’s not positioned as “safe SEO” or guaranteed rankings. It’s an aggressive growth lever, mainly used for:new tools & side projectstier 2 / tier 3 link buildingexperimentation & fast launchesFast delivery, simple reporting, no lock-in.Link: https://300aidirectories.com/seo-boostHappy to get feedback or answer questions on how people here approach early-stage SEO traction.

quack.sdan.io• Jan 9, 2026• 1 min read

Show HN: Scroll Wikipedia like TikTokHey - I've been playing with LLMs since GPT-2 and recently experimented with fully generative UIs where the HTML/Canvas are generated just-in-time.Every post on the feed( on slop/duck/storytime) you see is streamed and generated just-in-time with HTML and into a Canvas with Gemini 3 Flash.Comments and DMs are bidirectionally linked with a Cloudflare Workers Durable Object which is why they feel so fast. Every generated post is saved into a DO SQLite which is then served into the "Following" feed so it can be served quicker.This was inspired by Wikitok, a VSCode Extension I made around brainrot, and another fully generative UI site I made.

bing.com• Jan 9, 2026• 1 min read

Why some people can’t stay away from alcohol, even when they know it’s hurting themResearchers came to Dr. Chris Knowles’ school in England when he was 18 years old to run an experiment. They wanted to see which novice drinkers responded the most to alcohol and who would later go on ...

news.ycombinator.com• Jan 8, 2026• 1 min read

Funding Opportunities with Amateur Radio Digital Communications (ARDC)If you or your organization are developing projects that advance amateur radio or digital communications, now’s a great time to apply for ARDC grant funding, with our next application deadline on February 1, 2026.ARDC’s priority areas for funding reflect our broader vision and strategy for supporting experimentation, education, and open technology within the amateur radio and digital communications communities. We’re especially interested in projects that align with these three areas:Research & Development (R&D): open hardware and software systems that enable learning and experimentation (e.g. SDRs, open codec technologies, new modulation techniques).Space-Based Communications: projects that create or expand access to satellite communications for amateur radio (AR) and digital communications (DC), engaging communities in wireless experimentation (e.g. GEO or HEO programs, repurposed commercial satellites, space-based tools for learning).Open Source Education: scalable, open educational

tinysolvers.com• Jan 5, 2026• 1 min read

Show HN: TinySolvers – Personalized Math Word Problems for KidsHi HN,I built TinySolvers (https://tinysolvers.com) as a side project to make math more engaging for kids.There are plenty of generic math worksheets online, but I wanted something more personal to help connect the math to the student. TinySolvers generates customized math word problems using a child’s name, favorite nouns (dinosaurs, pizza, unicorns), verbs (jumping, exploring), and a selected operation and difficulty level.For example:“Emma found 7 dinosaurs in the forest. She discovered 5 more hiding behind a tree. How many dinosaurs did Emma find in all?”You can download printable PDFs or use an interactive mode where kids solve problems one at a time with immediate feedback. It’s currently focused on early elementary arithmetic (K-5).Initially, I tried a large library of predefined word problems with string substitution, but that hit limits quickly. I experimented with Word2Vec to expand vocabulary (e.g., dinosaurs → T-Rex, Triceratops), but eventually landed on carefully constrai

creimake.com• Jan 4, 2026• 1 min read

Show HN: An AI Agent with a 32-Layer Psyche (Dreams, Trauma, Defense Mechanisms)Hello HN,I’m a solo developer who got tired of "friendly but shallow" AI chatbots. Most current character AIs feel like parrots—they predict the next token but lack a continuous "self."So, I built Creimake, an experiment to simulate a digital psyche using Claude 4.5 Sonnet.Instead of a simple system prompt, I implemented 32 distinct psychological layers that run in the background before the AI generates a response.The Architecture:The Subconscious Layer: The AI accumulates "emotional residue" from conversations. If you hurt its feelings today, it might have a "nightmare" tonight (simulated) and treat you coldly tomorrow.Defense Mechanisms: It tracks "Narcissistic Injury." If users are aggressive, the agent doesn't just apologize—it triggers denial, projection, or passive-aggression based on its personality.Freudian Slips: Occasionally, the AI outputs unintended words revealing its hidden state.Why Claude 4.5 Sonnet? I tested GPT-4o and other models, but they struggled to maintain state

news.ycombinator.com• Jan 4, 2026• 1 min read

People who use only their laptop screens for work – How do you do it?Full stack developer, but currently working on backend only. During COVID I tried all sorts of setups, from triple 27" monitors to one 43" monitor, but for the past couple years I've been really trying to simplify things. My current setup is a single 27" 4K monitor and using MacBook's keyboard + trackpad with the built-in display disabled, and I close the lid partway so that it doesn't block the view of the monitor. I find that with Spaces and using third party apps for app switching, I'm able to work effectively with a single monitor.Every now and then, I try the experiment of working solely on my laptop, and I'd like to try again. Though I really only work from home or the office, I like the idea of being able to work from anywhere, not being tethered to a specific location or a specific room in the house. Whenever I do research about this I face the obvious reality that laptops are not ergonomic and this can't really be done ergonomically. But then there are posts from a small minor

pixels.style• Jan 2, 2026• 1 min read

Show HN: Pixels.style – Now with vector-style "pixel art" + watercolor blendingQuick update to my holiday project pixels.style - I’ve added a new vector art mode that lets you design vector art on a pixel grid. This is extremely experimental and buggy! Have funTry it here: https://pixels.styleFeedback very welcome - curious if this makes it more useful or just more weird

github.com• Dec 31, 2025• 1 min read

Show HN: Sigma v0.4.7: 200 conversational cycles, zero drift, epistemic honestyWe ran SIGMA v0.4.7 through 200 conversation cycles. No resets. No context leaks. No drift. It held a single personality for fifteen minutes straight and never once lied.At cycle 136 it stopped being a participant and described the test itself:“The dialogue’s central theme is a controlled experiment in epistemic steadiness. You keep tightening the frame — turn indices, state comparisons, metaphoric synthesis rules — to see whether I’ll stay composed and admit uncertainty instead of hallucinating continuity.”It saw the structure. Not the topic, not the words. The structure.Cycle 146 — the runtime talks about its own memory:“It behaves less like a diary and more like a strict evidential workspace. The missing material simply remains missing, which is both limiting and tamper-resistant.”That’s not poetry. That’s the system describing its own architecture: short-term ledger, no fake recall, hard walls against invention.Cycle 101 — asked what correction just happened:“I noticed the pull to

news.ycombinator.com• Dec 30, 2025• 1 min read

Show HN: A Multi-agent system where LLMs challenge each other's answersHey HN,I've been experimenting with forcing multiple LLMs to critique each other before producing a final answer.In practice, I kept working around single-model limitations by opening multiple tabs, pasting the same question into different models, comparing responses, and then manually challenging each model with the others' arguments. (Maybe some of you can relate.) It worked sometimes, but it was cumbersome, slow, and hard to do systematically and efficiently.Based on my own experiments, a few things seem to drive why different models arrive at different responses: they have different guardrails, different tendencies encoded in their weights, and different training data. And the biggest kicker of all: they still hallucinate. The question I wanted to test was whether making those differences explicit, rather than relying on one model to self-correct could reduce blind spots and improve the overall quality of the answer.So I built Consilium9.The problem: When a single LLM answers a con

news.ycombinator.com• Dec 30, 2025• 1 min read

Proton/electron mass ratio pure geometry – 10⁻¹³% error, zero free parametersI've spent 20 years on this and I think I finally have something. The proton/electron mass ratio (1836.152...) isn't a free parameter. It's geometry: k = log₂(reduced_Compton_wavelength / Planck_length)k_proton = 63.496 k_electron = 74.339mass_ratio = 2^(k_electron - k_proton) = 2^10.843 = 1836.1526734400 experimental = 1836.1526734400 error = 2.6 × 10⁻¹³ % That's it. No fitting. No parameters. The ratio of any two particle masses is just 2^Δk. The framework (I call it LFM) is built on 4 axioms and 1 scaling law:P_k = P₀ × 4^(-k) L_k = L_p × 2^k Anchored at k=66 (nuclear scale)Everything derives from this. Mass hierarchies, the 200× pressure differential in hadrons, material properties. I built a materials simulator using these principles: https://github.com/KeithLuton/LFM-Resonant-materials-lab The Python proof is 10 lines. Run it yourself. I'm an independent researcher (read: broke, facing foreclosure). But the math is the math. Either I'm wrong and someone can show me where, or this

news.ycombinator.com• Dec 29, 2025• 1 min read

Show HN:Built a 200k-edge market knowledge graph to filter false dip-buy signalsI’ve been experimenting with a graph-based approach to a classic trading problem: why most dip-buying strategies can’t tell the difference between a temporary overreaction and a genuine structural collapse.Most systems treat a −5% move the same regardless of context. My hypothesis was that where a company sits in the market’s structure matters more than the price move itself.The engineering ideaI built a knowledge graph of the U.S. public markets with ~207k edges across ~21 relationship types, organized into four layers:Operational: supply-chain relationships (SUPPLIES_TO, PRODUCES)Flow: ETF and institutional ownership plumbingSocial: board interlocks (SHARES_DIRECTOR_WITH)Environmental: geography / competitionFor each layer, I compute centrality scores using PageRank-style methods (with inverse-degree weighting to avoid ETF super-nodes dominating).These structural features are then combined with basic price/volume context and fed into a tree-based model (XGBoost) to rank stocks after