{"audienceFit":["Developers, builders, artists, and AI operators who want Kyanite-built tools working in their environment.","Teams who want the outcome from Kyanite open-source tools without doing every setup, adaptation, and handoff step alone.","People following the build notes, learning process, experiments, and public product work behind KyaniteLabs."],"blogPosts":[{"description":"The measured nmax2 acceptance inversion: 72.2% versus 22.7% for the draft-12 control. Throughput and engine comparisons, with single-lab limits.","primaryKeyword":"Play-1 speculative decoding acceptance","spanishUrl":"https://kyanitelabs.tech/es/blog/play1-numbers","title":"Play-1: the acceptance-inversion result","url":"https://kyanitelabs.tech/blog/play1-numbers"},{"description":"Signed permissions and hash-chained receipts catch real bypasses. Two acceptance-path defects were still open in the source report.","primaryKeyword":"adversarial custody chains research evidence","spanishUrl":"https://kyanitelabs.tech/es/blog/adversarial-custody-chains","title":"Adversarial custody chains: keep the attacks in the record","url":"https://kyanitelabs.tech/blog/adversarial-custody-chains"},{"description":"Measured sampler costs, guidance timing and model reload overhead. Quality judgments remain single-judge observations, not a leaderboard.","primaryKeyword":"local image generation consumer APU measurements","spanishUrl":"https://kyanitelabs.tech/es/blog/local-image-stack-measured","title":"Local image generation, measured on a consumer APU","url":"https://kyanitelabs.tech/blog/local-image-stack-measured"},{"description":"A routing policy for the actual GPT-6 lineup: Sol for discovery and coding, Luna for bounded high-volume processing, Astra when nothing else holds \u2014 each with a verification contract, and the lanes where a local 27B still beats all three.","primaryKeyword":"GPT-6 Sol vs Luna routing for agents","spanishUrl":"https://kyanitelabs.tech/es/blog/gpt-6-sol-luna-astra-routing-guide","title":"GPT-6 Sol vs. Luna vs. Astra: a routing policy for coding agents \u2014 and when local still wins","url":"https://kyanitelabs.tech/blog/gpt-6-sol-luna-astra-routing-guide"},{"description":"We benchmarked our own thesis: tiny model + deterministic guardrail layer vs raw capability. The curve is textbook, the trim trap is real, and the bench indicted our own product before anyone else could.","primaryKeyword":"small model benchmark video editing MCP guardrails","spanishUrl":"https://kyanitelabs.tech/es/blog/equalizer-bench-small-models-video","title":"The Equalizer Bench: a 3B that can't write ffmpeg, the same 3B shipping video edits, and the bug our own benchmark caught","url":"https://kyanitelabs.tech/blog/equalizer-bench-small-models-video"},{"description":"Spec decoding crashed our fastest lane on day one. The bug was one missing line in our fork; the fix bought +17.7% and a surprise about draft quantization.","primaryKeyword":"speculative decoding llama.cpp draft model iGPU speedup","spanishUrl":"https://kyanitelabs.tech/es/blog/the-one-line-bug","title":"The One-Line Bug That Crashed Our Fast Lane: finding, fixing, and measuring a speculative-decoding crash on a $1,400 mini-PC","url":"https://kyanitelabs.tech/blog/the-one-line-bug"},{"description":"A 35B reasoning model now runs shoulder to shoulder with our daily 27B on one $1,400 box, at the same time. Every number paired, same problems, same machine. The failures are in here too.","primaryKeyword":"run two local llms on one mini pc paired benchmark","spanishUrl":"https://kyanitelabs.tech/es/blog/two-models-one-mini-pc-paired-numbers","title":"Two models, one $1,400 mini-PC: the paired numbers, failures included","url":"https://kyanitelabs.tech/blog/two-models-one-mini-pc-paired-numbers"},{"description":"The title is new; the work is old. Twelve years of enterprise deployments plus public, measured AI work. The honest path, artifacts included.","primaryKeyword":"how to become a forward deployed engineer without degree","spanishUrl":"https://kyanitelabs.tech/es/blog/how-i-became-a-forward-deployed-engineer","title":"How I became a forward deployed engineer without a software engineer title","url":"https://kyanitelabs.tech/blog/how-i-became-a-forward-deployed-engineer"},{"description":"The market says it cannot find people who can build AI evals. The skill is learnable and I published mine: 495 trials, certified floors, sabotage cell, open source.","primaryKeyword":"AI evaluation benchmark LLM evals","spanishUrl":"https://kyanitelabs.tech/es/blog/evals-are-the-fde-skill","title":"Evals are the FDE skill nobody lists: my 495-trial public benchmark","url":"https://kyanitelabs.tech/blog/evals-are-the-fde-skill"},{"description":"Four titles, one job family, different coding bars. A decoder that reads any posting and tells you what you are actually signing up for.","primaryKeyword":"forward deployed engineer vs solutions engineer","spanishUrl":"https://kyanitelabs.tech/es/blog/fde-title-decoder","title":"Forward deployed vs solutions engineer vs implementation vs customer engineer: the title decoder","url":"https://kyanitelabs.tech/blog/fde-title-decoder"},{"description":"The straight answer, then the receipts: the whole FDE job run end to end on a $1,400 mini-PC, with public evals.","primaryKeyword":"what does a forward deployed engineer do","spanishUrl":"https://kyanitelabs.tech/es/blog/what-does-a-forward-deployed-engineer-do","title":"What does a forward deployed engineer actually do? A demo on a $1,400 mini-PC","url":"https://kyanitelabs.tech/blog/what-does-a-forward-deployed-engineer-do"},{"description":"Not is-it-smart but can-you-hand-it-work-and-walk-away. 495 certified trials, then re-validated at deeper n after the product changed: 965 total, floors to 92.8%.","primaryKeyword":"local LLM delegation benchmark certified floors Strix Halo","spanishUrl":"https://kyanitelabs.tech/es/blog/the-delegation-card","title":"The Delegation Card: we asked a $1,400 mini-PC to take our jobs","url":"https://kyanitelabs.tech/blog/the-delegation-card"},{"description":"Every dial measured, every number public: the frozen optimal config for a 27B on a $1,400 mini-PC.","primaryKeyword":"Qwen 27B Strix Halo optimal serving config llama.cpp","spanishUrl":"https://kyanitelabs.tech/es/blog/qwen-27b-strix-halo-complete","title":"Qwen 3.8 27B on Strix Halo: the complete measured story","url":"https://kyanitelabs.tech/blog/qwen-27b-strix-halo-complete"},{"description":"A methods note on reasoning-effort calibration. Thinking rescued 15/40 hards vs 4/40 off. On easy tasks it bought nothing. Publish the knee.","primaryKeyword":"reasoning effort calibration knee Qwen3.8-27B","spanishUrl":"https://kyanitelabs.tech/es/blog/lab-notes-measured-knees","title":"Lab Notes: the measured-knees method for reasoning effort","url":"https://kyanitelabs.tech/blog/lab-notes-measured-knees"},{"description":"20/30 = 67% LiveCodeBench-30 on a $1,400 mini-PC. Wilson 95% CI 49-81%. Easy 10/10, medium 8/10, hard 2/10. n=30 public subset. Not the card.","primaryKeyword":"LiveCodeBench Qwen3.8-27B Q4_K_XL Strix Halo","spanishUrl":"https://kyanitelabs.tech/es/blog/lab-notes-livecodebench-30","title":"Lab Notes: 67% LiveCodeBench-30 on a $1,400 rig","url":"https://kyanitelabs.tech/blog/lab-notes-livecodebench-30"},{"description":"We found a llama.cpp regression and reverted it. n=6 battery on a $1,400 rig: 0/6 slashes before, 5/6 after. Same Q4_K_XL. Not a quant story.","primaryKeyword":"llama.cpp c7d8722 HIP vision NaN logits","spanishUrl":"https://kyanitelabs.tech/es/blog/lab-notes-llamacpp-revert","title":"Lab Notes: we reverted a llama.cpp regression","url":"https://kyanitelabs.tech/blog/lab-notes-llamacpp-revert"},{"description":"28/30 = 93% HumanEval on a $1,400 mini-PC. Qwen3.8-27B Q4_K_XL. Temp 0, thinking off. Failures: 50 and 145. Raw log linked.","primaryKeyword":"HumanEval Qwen3.8-27B Q4_K_XL Strix Halo","spanishUrl":"https://kyanitelabs.tech/es/blog/lab-notes-humaneval-93","title":"Lab Notes: 93% HumanEval on a $1,400 rig","url":"https://kyanitelabs.tech/blog/lab-notes-humaneval-93"},{"description":"The number first: this model is 75% not a transformer. 48 of 64 layers keep a running state. On a $1,400 rig, load 198k once (1818s), then query in 9-27s.","primaryKeyword":"Gated DeltaNet Qwen3.8-27B 262k context","spanishUrl":"https://kyanitelabs.tech/es/blog/lab-notes-cant-forget","title":"Lab Notes: the model that can't forget but can't remember","url":"https://kyanitelabs.tech/blog/lab-notes-cant-forget"},{"description":"We published a basin. The serving binary was the hole. After the c7d8722 revert: 6/6 HIT at 198k. n=1 map of the fixed build.","primaryKeyword":"llama.cpp c7d8722 deep context needle retrieval","spanishUrl":"https://kyanitelabs.tech/es/blog/lab-notes-degenerate-basin","title":"Lab Notes: the basin was a bug","url":"https://kyanitelabs.tech/blog/lab-notes-degenerate-basin"},{"description":"The q8-to-q4 KV question from night 1 got its paired answer: same accuracy, zero tripwires, half the cache. Plus the label we had to correct in public when the server's counter beat our estimate.","primaryKeyword":"q4 KV cache quantization llama.cpp","spanishUrl":"https://kyanitelabs.tech/es/blog/lab-notes-the-kv-verdict","title":"Lab Notes: the KV verdict","url":"https://kyanitelabs.tech/blog/lab-notes-the-kv-verdict"},{"description":"A $1,400 mini-PC serving a 27B at 262k context asks one question: does capping thinking cost accuracy? The paired answer, the invalid first attempt it survived, and the doctrine that followed.","primaryKeyword":"LLM thinking budget cap accuracy","spanishUrl":"https://kyanitelabs.tech/es/blog/lab-notes-verdict-night","title":"Lab Notes: verdict night","url":"https://kyanitelabs.tech/blog/lab-notes-verdict-night"},{"description":"A fully measured night-and-evening of tuning a 27B dense model on a Strix Halo mini-PC: the honest bands, the reversal, the crash doctrine, and the EC detective story.","primaryKeyword":"Qwen 27B Strix Halo local LLM tuning","spanishUrl":"https://kyanitelabs.tech/es/blog/qwen-27b-strix-halo-one-week-local","title":"One mini-PC, one night, and the numbers that argued with themselves: tuning Qwen3.8-27B on Strix Halo","url":"https://kyanitelabs.tech/blog/qwen-27b-strix-halo-one-week-local"},{"description":"The useful agent pattern is not a prettier prompt. It is a tool surface the agent can call, inspect, verify, and revise.","primaryKeyword":"verifiable AI agent tools","spanishUrl":"https://kyanitelabs.tech/es/blog/agents-need-verifiable-tools","title":"Agents need verifiable tools, not better prompt theater","url":"https://kyanitelabs.tech/blog/agents-need-verifiable-tools"},{"description":"A repo is not just storage. It is evidence of decisions, repairs, release behavior, naming drift, test gaps, and what the builder actually knows how to finish.","primaryKeyword":"repo archaeology for AI teams","spanishUrl":"https://kyanitelabs.tech/es/blog/repo-history-is-a-product-signal","title":"Repo history is a product signal","url":"https://kyanitelabs.tech/blog/repo-history-is-a-product-signal"},{"description":"A useful open-source tool still needs a path from public repo to working environment. That path is product work, not an afterthought.","primaryKeyword":"AI tool implementation support","spanishUrl":"https://kyanitelabs.tech/es/blog/implementation-help-is-product-surface","title":"Implementation help is part of the product surface","url":"https://kyanitelabs.tech/blog/implementation-help-is-product-surface"},{"description":"Kinocut gives AI agents callable handles on timelines, effects, Hyperframes, and finished media at kinocut.dev.","primaryKeyword":"video editing MCP server","spanishUrl":"https://kyanitelabs.tech/es/blog/why-kinocut-matters","title":"Why Kinocut matters","url":"https://kyanitelabs.tech/blog/why-kinocut-matters"},{"description":"The argument behind the video: output quality is not just probability. It is architecture, filters, and human taste.","primaryKeyword":"agentic systems","spanishUrl":"https://kyanitelabs.tech/es/blog/infinite-monkey-agentic-systems","title":"Infinite monkeys, LLMs, and the room around the machine","url":"https://kyanitelabs.tech/blog/infinite-monkey-agentic-systems"},{"description":"A practical checklist for turning a working tool, workflow, or rough app into something other people can understand, install, and use.","primaryKeyword":"AI tool implementation","spanishUrl":"https://kyanitelabs.tech/es/blog/ai-tool-implementation-checklist","title":"What a working AI tool needs before people can use it","url":"https://kyanitelabs.tech/blog/ai-tool-implementation-checklist"},{"description":"The checklist Kyanite uses to decide whether an MCP server is a toy, a usable tool, or something worth implementing.","primaryKeyword":"MCP server implementation","spanishUrl":"https://kyanitelabs.tech/es/blog/mcp-server-implementation-checklist","title":"MCP server implementation checklist","url":"https://kyanitelabs.tech/blog/mcp-server-implementation-checklist"},{"description":"Why commit history is one of the strongest proof sources for learning diagnostics, implementation help, and engineering trust.","primaryKeyword":"repo archaeology","spanishUrl":"https://kyanitelabs.tech/es/blog/repo-archaeology-proof-assets","title":"Repo archaeology turns history into proof","url":"https://kyanitelabs.tech/blog/repo-archaeology-proof-assets"},{"description":"What Kyanite adds so search engines and AI assistants can understand the tools, products, proof, and support path.","primaryKeyword":"AI discovery","spanishUrl":"https://kyanitelabs.tech/es/blog/ai-discovery-llms-txt-geo","title":"AI discovery needs more than a sitemap","url":"https://kyanitelabs.tech/blog/ai-discovery-llms-txt-geo"}],"contact":"info@kyanitelabs.tech","founder":{"name":"Simon Gonzalez de Cruz","spanishUrl":"https://kyanitelabs.tech/es/about","url":"https://kyanitelabs.tech/about"},"magnets":[{"description":"Index of KyaniteLabs worksheets, checklists, settings cards, and model-selection references.","title":"Free Resources","url":"https://kyanitelabs.tech/magnets"},{"description":"A practical local-model settings reference.","title":"Qwen3.8-27B Settings Card","url":"https://kyanitelabs.tech/magnets/settings-card"},{"description":"A worksheet for splitting agent work into bounded, verifiable lanes.","title":"Agent Delegation Worksheet","url":"https://kyanitelabs.tech/magnets/delegation-worksheet"},{"description":"A checklist for designing evaluations with evidence and failure controls.","title":"Evaluation Design Checklist","url":"https://kyanitelabs.tech/magnets/eval-checklist"},{"description":"A snapshot-based model-selection reference for intelligence, cost, and speed.","title":"LLM Comparison Cheat Sheet","url":"https://kyanitelabs.tech/magnets/llm-cheat-sheet"}],"paidPaths":[{"deliverables":["Get Kyanite tools installed, configured, and checked","Adapt an open-source tool to the real workflow and tradeoffs that matter","Explain setup, architecture, docs, and handoff","Build implementation scope for MCP, media, localization, or dev tooling work"],"name":"Implementation and Advising","url":"https://kyanitelabs.tech/implementation"},{"deliverables":["MCP server, CLI, or Python/TypeScript package","Tool schemas, examples, tests, docs, and install path","Public or internal implementation surface"],"name":"MCP / Agent Tool Build","url":"https://kyanitelabs.tech/#support"},{"deliverables":["Kinocut-backed media workflow","Repeatable video assembly recipes","Clips, captions, posting assets, and review handoff"],"name":"Agentic Media System","url":"https://kyanitelabs.tech/#support"},{"deliverables":["DialectOS-backed Spanish localization QA","Dialect, register, glossary, and structure checks","Implementation-ready localization report"],"name":"Spanish Launch QA","url":"https://kyanitelabs.tech/#support"}],"products":[],"publicRepositories":[{"description":"Guardrailed video editing for AI agents: MCP server, kino CLI, Video Receipts, and Hyperframes.","image":"/static/brand/projects/kyanite-project-kinocut-1672x941.webp","language":"Python","name":"Kinocut","proof_role":"The flagship proof that agents can operate timelines, effects, and repeatable media pipelines.","repo":"https://github.com/KyaniteLabs/kinocut","tag":"Media MCP","tile_code":"KC","updated":"2026-09-01","url":"https://kinocut.dev"},{"description":"Design review skills and recorded evidence for typography, color, spacing, accessibility, interface states, and frontend ship decisions. Judgments are scoped assessments, not an effectiveness guarantee.","language":"Markdown","name":"tastecheck","proof_role":"The public landing page shows the review workflow and example systems; release evidence and effectiveness are separate claims.","repo":"https://github.com/KyaniteLabs/tastecheck","tag":"Frontend Craft","tile_code":"TC","updated":"2026-09-04","url":"https://kyanitelabs.github.io/tastecheck/"},{"description":"Time estimation MCP server for PERT, COCOMO II, Monte Carlo, sprint forecasting, token-to-time mapping, cost estimation, and schedule risk tools.","image":"/static/brand/projects/kyanite-project-epoch-1672x941.webp","language":"TypeScript","name":"Epoch","proof_role":"Turns planning uncertainty into agent-callable forecasting tools.","repo":"https://github.com/KyaniteLabs/Epoch","tag":"Estimation MCP","tile_code":"EP","updated":"2026-09-02","url":"https://kyanitelabs.github.io/Epoch/"},{"description":"Local completion-evidence review for AI-built apps: read-only audits, verifier challenges, bounded scores, guided fixes, learning plans, CLI, and MCP. Scores do not certify production safety.","image":"/static/brand/projects/kyanite-project-checkyourself-1672x941.webp","language":"Python","name":"checkyourself","proof_role":"Makes quality, security, and launch risk inspectable before an AI-built app ships.","repo":"https://github.com/KyaniteLabs/checkyourself","tag":"Readiness Audit","tile_code":"CY","updated":"2026-09-04","url":"https://kyanitelabs.github.io/checkyourself/"},{"description":"Spanish dialect localization MCP server and CLI across 25 regional variants with register control, structure preservation, and QA gates.","image":"/static/brand/projects/kyanite-project-dialectos-1672x941.webp","language":"TypeScript","name":"DialectOS","proof_role":"Makes Spanish launch quality inspectable instead of treating localization as generic translation.","tag":"Localization MCP","tile_code":"DX","updated":"2026-09-04","url":"https://github.com/KyaniteLabs/DialectOS"},{"description":"An honest index of the Resonant gift line: 19 open-source gifts built for the ResonantOS ecosystem \u2014 measurement benches, security tooling, local-service add-ons, and receipted knowledge.","language":null,"name":"resonant-gifts","proof_role":"Indexes the full line \u2014 every gift public, licensed, and independently reviewed.","tag":"ResonantOS Ecosystem","tile_code":"RG","updated":"2026-09-04","url":"https://github.com/simongonzalezdc/resonant-gifts"},{"description":"Injection defense + contest for the ResonantOS ecosystem: the doctrine, a labeled corpus of known public attack classes on synthetic fixtures, and a stdlib-only reference checker.","language":"Python","name":"resonant-constable","proof_role":"The defense discipline behind a merged upstream security fix, made reusable.","tag":"ResonantOS Ecosystem","tile_code":"IC","updated":"2026-09-03","url":"https://github.com/simongonzalezdc/resonant-constable"},{"description":"The ResonantDAO whitepaper's contextual-authority mechanism as a self-contained add-on: weight from verified outcomes in exactly the dimensions a question declares salient.","language":"Shell","name":"resonant-authority","proof_role":"The whitepaper's governance mechanism as working code \u2014 non-canonical, MIT.","tag":"ResonantOS Ecosystem","tile_code":"RA","updated":"2026-09-02","url":"https://github.com/simongonzalezdc/resonant-authority"},{"description":"AI creative coding studio for p5.js, GLSL, Three.js, music, video, and model-agnostic generative art workflows.","image":"/static/brand/projects/kyanite-project-liminal-1672x941.webp","language":"TypeScript","name":"liminal","proof_role":"Shows Kyanite can build creative tools where agents touch code, shaders, media, and taste.","tag":"Creative Coding","tile_code":"LM","updated":"2026-09-01","url":"https://github.com/KyaniteLabs/liminal"},{"description":"Site surfaces and public faces for the Liminal / Sinter stack, focused on shipping Liminal product sites.","image":"/static/brand/projects/kyanite-project-liminal-sites-1672x941.webp","language":"TypeScript","name":"liminal-sites","proof_role":"Proves the website itself can evolve through constrained, inspectable design systems.","tag":"Living Websites","tile_code":"LS","updated":"2026-09-02","url":"https://github.com/KyaniteLabs/liminal-sites"},{"description":"Local-first pattern synthesis engine for identity, brand voice, design systems, naming research, and lens-specific outputs.","image":"/static/brand/projects/kyanite-project-elixis-1672x941.webp","language":"Python","name":"Elixis","proof_role":"Turns fuzzy identity and naming work into source-backed synthesis.","tag":"Pattern Synthesis","tile_code":"EX","updated":"2026-09-02","url":"https://github.com/KyaniteLabs/Elixis"},{"description":"Personal growth OS in TypeScript for journaling, emotional check-ins, habits, goals, tasks, sleep logs, decluttering, and self-awareness workflows.","image":"/static/brand/projects/kyanite-project-innerscape-1672x941.webp","language":"TypeScript","name":"Innerscape","proof_role":"Shows the same build-and-implementation pattern applied to intimate, data-rich workflows.","tag":"Personal OS","tile_code":"IN","updated":"2026-09-02","url":"https://github.com/KyaniteLabs/Innerscape"},{"description":"Git repository archaeology framework for mining commit history, detecting signals, running 6 analysis vectors, and generating engineering reports.","image":"/static/brand/projects/kyanite-project-devarch-framework-1672x941.webp","language":"Python","name":"devarch-framework","proof_role":"Shows Kyanite can turn development history into evidence, not vibes.","tag":"Repo Intelligence","tile_code":"DF","updated":"2026-09-02","url":"https://github.com/KyaniteLabs/devarch-framework"},{"description":"Forensic git-history learning diagnostic for AI-assisted developers that turns commit history into evidence-backed study plans and HTML reports.","image":"/static/brand/projects/kyanite-project-dev-learning-archaeologist-1672x941.webp","language":"JavaScript","name":"Dev Learning Archaeologist","proof_role":"Turns repo behavior into a readable diagnostic artifact.","tag":"Learning Diagnostics","tile_code":"DA","updated":"2026-09-02","url":"https://github.com/simongonzalezdc/dev-learning-archaeologist"},{"description":"Folder-based AI food-memory researcher for investigating half-remembered family dishes from sound-alikes, smells, colors, and cultural memory.","image":"/static/brand/projects/kyanite-project-achiote-food-memory-1672x941.webp","language":"Claude Project","name":"achiote-food-memory-researcher","proof_role":"Extends Kyanite's evidence style into cultural memory and research assistants.","tag":"Food Memory","tile_code":"AC","updated":"2026-09-02","url":"https://github.com/simongonzalezdc/achiote-food-memory-researcher"},{"description":"Executive-function accessibility coach that turns a messy stuck point into one humane next move.","image":"/static/brand/projects/kyanite-project-unstuck-coach-1672x941.webp","language":"CSS","name":"unstuck-coach","proof_role":"Applies agentic tooling to accessibility, pacing, and human next-step design.","tag":"Executive Function","tile_code":"UC","updated":"2026-09-01","url":"https://github.com/simongonzalezdc/unstuck-coach"},{"description":"Portfolio prototype for asset-heavy field-service operators: equipment records, service notes, deficiencies, scheduling, and billing handoffs.","image":"/static/brand/projects/kyanite-project-tradesflow-1672x941.webp","language":"TypeScript","name":"tradesflow","proof_role":"Shows implementation thinking for real-world operators with assets, visits, and billing handoffs.","tag":"Field Operations","tile_code":"TF","updated":"2026-09-01","url":"https://github.com/simongonzalezdc/tradesflow"},{"description":"FastAPI health advocacy web app for patient tooling, local LLM exploration, and medical NLP workflows.","image":"/static/brand/projects/kyanite-project-healthadvocate-1672x941.webp","language":"Python","name":"healthadvocate","proof_role":"Explores private, local-first health tooling without treating care context as generic chat.","tag":"Health Advocacy","tile_code":"HA","updated":"2026-09-02","url":"https://github.com/simongonzalezdc/healthadvocate"},{"description":"Free open-source ceramic glaze calculator, UMF analyzer, CTE estimator, recipe manager, and studio tool for potters and ceramic artists.","image":"/static/brand/projects/kyanite-project-openglaze-1672x941.webp","language":"Python","name":"openglaze","proof_role":"Proves Kyanite can ship useful software outside the generic AI-tool bubble.","tag":"Domain Software","tile_code":"OG","updated":"2026-09-02","url":"https://github.com/KyaniteLabs/openglaze"}],"site":{"description":"Open-source proof and paid implementation help for getting AI tools, MCP systems, media pipelines, localization QA, and repo diagnostics working in real environments.","discovery":{"llms":"https://kyanitelabs.tech/llms.txt","llmsFull":"https://kyanitelabs.tech/llms-full.txt","rss":"https://kyanitelabs.tech/feed.xml","sitemap":"https://kyanitelabs.tech/sitemap.xml"},"languages":["en","es"],"name":"KyaniteLabs","parentOrganization":{"name":"PuenteWorks","role":"Legal and business container for broader consulting, workflow, and AI operations work.","url":"https://puenteworks.com"},"portfolioRelationship":{"kyaniteLabs":"https://kyanitelabs.tech","puenteWorks":"https://puenteworks.com","summary":"PuenteWorks leads business workflow and consulting engagements; KyaniteLabs is the technical product and implementation lab inside the same owned portfolio."},"url":"https://kyanitelabs.tech"},"spanishBlogPosts":[{"description":"The measured nmax2 acceptance inversion: 72.2% versus 22.7% for the draft-12 control. Throughput and engine comparisons, with single-lab limits.","englishUrl":"https://kyanitelabs.tech/blog/play1-numbers","primaryKeyword":"Play-1 speculative decoding acceptance","title":"Play-1: the acceptance-inversion result","url":"https://kyanitelabs.tech/es/blog/play1-numbers"},{"description":"Signed permissions and hash-chained receipts catch real bypasses. Two acceptance-path defects were still open in the source report.","englishUrl":"https://kyanitelabs.tech/blog/adversarial-custody-chains","primaryKeyword":"adversarial custody chains research evidence","title":"Adversarial custody chains: keep the attacks in the record","url":"https://kyanitelabs.tech/es/blog/adversarial-custody-chains"},{"description":"Measured sampler costs, guidance timing and model reload overhead. Quality judgments remain single-judge observations, not a leaderboard.","englishUrl":"https://kyanitelabs.tech/blog/local-image-stack-measured","primaryKeyword":"local image generation consumer APU measurements","title":"Local image generation, measured on a consumer APU","url":"https://kyanitelabs.tech/es/blog/local-image-stack-measured"},{"description":"A routing policy for the actual GPT-6 lineup: Sol for discovery and coding, Luna for bounded high-volume processing, Astra when nothing else holds \u2014 each with a verification contract, and the lanes where a local 27B still beats all three.","englishUrl":"https://kyanitelabs.tech/blog/gpt-6-sol-luna-astra-routing-guide","primaryKeyword":"GPT-6 Sol vs Luna routing for agents","title":"GPT-6 Sol vs. Luna vs. Astra: a routing policy for coding agents \u2014 and when local still wins","url":"https://kyanitelabs.tech/es/blog/gpt-6-sol-luna-astra-routing-guide"},{"description":"We benchmarked our own thesis: tiny model + deterministic guardrail layer vs raw capability. The curve is textbook, the trim trap is real, and the bench indicted our own product before anyone else could.","englishUrl":"https://kyanitelabs.tech/blog/equalizer-bench-small-models-video","primaryKeyword":"small model benchmark video editing MCP guardrails","title":"The Equalizer Bench: a 3B that can't write ffmpeg, the same 3B shipping video edits, and the bug our own benchmark caught","url":"https://kyanitelabs.tech/es/blog/equalizer-bench-small-models-video"},{"description":"Spec decoding crashed our fastest lane on day one. The bug was one missing line in our fork; the fix bought +17.7% and a surprise about draft quantization.","englishUrl":"https://kyanitelabs.tech/blog/the-one-line-bug","primaryKeyword":"speculative decoding llama.cpp draft model iGPU speedup","title":"The One-Line Bug That Crashed Our Fast Lane: finding, fixing, and measuring a speculative-decoding crash on a $1,400 mini-PC","url":"https://kyanitelabs.tech/es/blog/the-one-line-bug"},{"description":"Un modelo de razonamiento de 35B ya corre junto a nuestro 27B diario en una caja de $1,400, al mismo tiempo. Cada n\u00famero emparejado, mismos problemas, misma m\u00e1quina. Los fracasos tambi\u00e9n est\u00e1n aqu\u00ed.","englishUrl":"https://kyanitelabs.tech/blog/two-models-one-mini-pc-paired-numbers","primaryKeyword":"run two local llms on one mini pc paired benchmark","title":"Dos modelos, una mini-PC de $1,400: los n\u00fameros emparejados, fracasos incluidos","url":"https://kyanitelabs.tech/es/blog/two-models-one-mini-pc-paired-numbers"},{"description":"The title is new; the work is old. Twelve years of enterprise deployments plus public, measured AI work. The honest path, artifacts included.","englishUrl":"https://kyanitelabs.tech/blog/how-i-became-a-forward-deployed-engineer","primaryKeyword":"how to become a forward deployed engineer without degree","title":"How I became a forward deployed engineer without a software engineer title","url":"https://kyanitelabs.tech/es/blog/how-i-became-a-forward-deployed-engineer"},{"description":"The market says it cannot find people who can build AI evals. The skill is learnable and I published mine: 495 trials, certified floors, sabotage cell, open source.","englishUrl":"https://kyanitelabs.tech/blog/evals-are-the-fde-skill","primaryKeyword":"AI evaluation benchmark LLM evals","title":"Evals are the FDE skill nobody lists: my 495-trial public benchmark","url":"https://kyanitelabs.tech/es/blog/evals-are-the-fde-skill"},{"description":"Cuatro t\u00edtulos, una familia de trabajo, barras de c\u00f3digo distintas. Un decodificador para leer cualquier vacante y saber en qu\u00e9 te est\u00e1s metiendo.","englishUrl":"https://kyanitelabs.tech/blog/fde-title-decoder","primaryKeyword":"forward deployed engineer vs solutions engineer","title":"Forward deployed vs solutions engineer vs implementation vs customer engineer: el decodificador","url":"https://kyanitelabs.tech/es/blog/fde-title-decoder"},{"description":"La respuesta directa y luego los recibos: todo el trabajo de un FDE sobre un mini-PC de $1.400, con evals p\u00fablicas.","englishUrl":"https://kyanitelabs.tech/blog/what-does-a-forward-deployed-engineer-do","primaryKeyword":"que hace un forward deployed engineer","title":"Qu\u00e9 hace realmente un forward deployed engineer","url":"https://kyanitelabs.tech/es/blog/what-does-a-forward-deployed-engineer-do"},{"description":"Not is-it-smart but can-you-hand-it-work-and-walk-away. 495 certified trials, then re-validated at deeper n after the product changed: 965 total, floors to 92.8%.","englishUrl":"https://kyanitelabs.tech/blog/the-delegation-card","primaryKeyword":"local LLM delegation benchmark certified floors Strix Halo","title":"The Delegation Card: we asked a $1,400 mini-PC to take our jobs","url":"https://kyanitelabs.tech/es/blog/the-delegation-card"},{"description":"Cada dial medido, cada numero publico: la configuracion optima congelada para un 27B en un mini-PC de $1.400.","englishUrl":"https://kyanitelabs.tech/blog/qwen-27b-strix-halo-complete","primaryKeyword":"Qwen 27B Strix Halo configuracion optima de serving llama.cpp","title":"Qwen 3.8 27B en Strix Halo: la historia completa, medida","url":"https://kyanitelabs.tech/es/blog/qwen-27b-strix-halo-complete"},{"description":"A methods note on reasoning-effort calibration. Thinking rescued 15/40 hards vs 4/40 off. On easy tasks it bought nothing. Publish the knee.","englishUrl":"https://kyanitelabs.tech/blog/lab-notes-measured-knees","primaryKeyword":"reasoning effort calibration knee Qwen3.8-27B","title":"Lab Notes: the measured-knees method for reasoning effort","url":"https://kyanitelabs.tech/es/blog/lab-notes-measured-knees"},{"description":"20/30 = 67% LiveCodeBench-30 en un mini-PC de $1.400. IC Wilson 95% 49-81%. Easy 10/10, medium 8/10, hard 2/10. Subset n=30. No es la card.","englishUrl":"https://kyanitelabs.tech/blog/lab-notes-livecodebench-30","primaryKeyword":"LiveCodeBench Qwen3.8-27B Q4_K_XL Strix Halo","title":"Notas de lab: 67% LiveCodeBench-30 en un rig de $1.400","url":"https://kyanitelabs.tech/es/blog/lab-notes-livecodebench-30"},{"description":"Encontramos una regresi\u00f3n de llama.cpp y la revertimos. Bater\u00eda n=6 en un rig de $1.400: 0/6 slashes antes, 5/6 despu\u00e9s. Mismo Q4_K_XL. No es una historia de quant.","englishUrl":"https://kyanitelabs.tech/blog/lab-notes-llamacpp-revert","primaryKeyword":"llama.cpp c7d8722 HIP visi\u00f3n NaN logits","title":"Notas de lab: revertimos una regresi\u00f3n de llama.cpp","url":"https://kyanitelabs.tech/es/blog/lab-notes-llamacpp-revert"},{"description":"28/30 = 93% HumanEval en un mini-PC de $1.400. Qwen3.8-27B Q4_K_XL. Temp 0, thinking off. Fallos: 50 y 145. Log crudo linkeado.","englishUrl":"https://kyanitelabs.tech/blog/lab-notes-humaneval-93","primaryKeyword":"HumanEval Qwen3.8-27B Q4_K_XL Strix Halo","title":"Notas de lab: 93% HumanEval en un rig de $1.400","url":"https://kyanitelabs.tech/es/blog/lab-notes-humaneval-93"},{"description":"El n\u00famero primero: este modelo es 75% no transformer. 48 de 64 capas guardan un estado. En un rig de $1.400, carga 198k una vez (1818s) y consulta en 9-27s.","englishUrl":"https://kyanitelabs.tech/blog/lab-notes-cant-forget","primaryKeyword":"Gated DeltaNet Qwen3.8-27B contexto 262k","title":"Notas de lab: el modelo que no puede olvidar pero no puede recordar","url":"https://kyanitelabs.tech/es/blog/lab-notes-cant-forget"},{"description":"Publicamos una cuenca. El hueco era el binario. Tras revertir c7d8722: 6/6 HIT a 198k. Mapa n=1 del build arreglado.","englishUrl":"https://kyanitelabs.tech/blog/lab-notes-degenerate-basin","primaryKeyword":"llama.cpp c7d8722 retrieval contexto profundo","title":"Notas de lab: la cuenca era un bug","url":"https://kyanitelabs.tech/es/blog/lab-notes-degenerate-basin"},{"description":"La pregunta q8-a-q4 del KV de la noche 1 ya tiene respuesta pareada: misma accuracy, cero tripwires, la mitad del cache. Y el label que tuvimos que corregir en p\u00fablico cuando el contador del server le gan\u00f3 al estimado.","englishUrl":"https://kyanitelabs.tech/blog/lab-notes-the-kv-verdict","primaryKeyword":"cuantizaci\u00f3n cache KV q4 llama.cpp","title":"Notas de lab: el veredicto del KV","url":"https://kyanitelabs.tech/es/blog/lab-notes-the-kv-verdict"},{"description":"Un mini-PC de $1.400 sirviendo un 27B a 262k de contexto hace una pregunta: \u00bfcapear cu\u00e1nto piensa el modelo te cuesta accuracy? La respuesta pareada, el primer intento inv\u00e1lido, y la doctrina que qued\u00f3.","englishUrl":"https://kyanitelabs.tech/blog/lab-notes-verdict-night","primaryKeyword":"tope de thinking LLM accuracy","title":"Notas de lab: la noche del veredicto","url":"https://kyanitelabs.tech/es/blog/lab-notes-verdict-night"},{"description":"Una noche y una tarde de medicion afinando un modelo denso de 27B en un mini-PC Strix Halo: las bandas honestas, la reversion, la doctrina de crashes y la historia detectivesca del EC.","englishUrl":"https://kyanitelabs.tech/blog/qwen-27b-strix-halo-one-week-local","primaryKeyword":"Qwen 27B Strix Halo LLM local","title":"Un mini-PC, una noche y los numeros que discutian entre si: afinando Qwen3.8-27B en Strix Halo","url":"https://kyanitelabs.tech/es/blog/qwen-27b-strix-halo-one-week-local"},{"description":"El patron util no es un prompt mas bonito. Es una superficie de herramienta que el agente puede llamar, inspeccionar, verificar y revisar.","englishUrl":"https://kyanitelabs.tech/blog/agents-need-verifiable-tools","primaryKeyword":"verifiable AI agent tools","title":"Los agentes necesitan herramientas verificables, no mejor teatro de prompts","url":"https://kyanitelabs.tech/es/blog/agents-need-verifiable-tools"},{"description":"Un repo no es solo almacenamiento. Es evidencia de decisiones, reparaciones, releases, cambios de nombre, huecos de pruebas y oficio real.","englishUrl":"https://kyanitelabs.tech/blog/repo-history-is-a-product-signal","primaryKeyword":"repo archaeology for AI teams","title":"El historial del repo es una se\u00f1al de producto","url":"https://kyanitelabs.tech/es/blog/repo-history-is-a-product-signal"},{"description":"Una herramienta open source util todavia necesita una ruta desde repo publico hasta entorno funcionando. Esa ruta es producto.","englishUrl":"https://kyanitelabs.tech/blog/implementation-help-is-product-surface","primaryKeyword":"AI tool implementation support","title":"La ayuda de implementacion es parte de la superficie del producto","url":"https://kyanitelabs.tech/es/blog/implementation-help-is-product-surface"},{"description":"Kinocut da a los agentes de IA herramientas llamables sobre timelines, efectos, Hyperframes y medios terminados en kinocut.dev.","englishUrl":"https://kyanitelabs.tech/blog/why-kinocut-matters","primaryKeyword":"video editing MCP server","title":"Por que importa Kinocut","url":"https://kyanitelabs.tech/es/blog/why-kinocut-matters"},{"description":"El argumento detras del video: la calidad no es solo probabilidad. Es arquitectura, filtros y gusto humano.","englishUrl":"https://kyanitelabs.tech/blog/infinite-monkey-agentic-systems","primaryKeyword":"agentic systems","title":"Monos infinitos, LLMs y el cuarto alrededor de la maquina","url":"https://kyanitelabs.tech/es/blog/infinite-monkey-agentic-systems"},{"description":"Una guia practica para convertir una herramienta, flujo o app medio cruda en algo que otros puedan entender, instalar y usar.","englishUrl":"https://kyanitelabs.tech/blog/ai-tool-implementation-checklist","primaryKeyword":"AI tool implementation","title":"Lo que una herramienta de IA necesita antes de que alguien pueda usarla","url":"https://kyanitelabs.tech/es/blog/ai-tool-implementation-checklist"},{"description":"El checklist que Kyanite usa para distinguir un servidor MCP de juguete, una herramienta usable y algo que vale la pena implementar.","englishUrl":"https://kyanitelabs.tech/blog/mcp-server-implementation-checklist","primaryKeyword":"MCP server implementation","title":"Checklist de implementacion para servidores MCP","url":"https://kyanitelabs.tech/es/blog/mcp-server-implementation-checklist"},{"description":"Por que el historial de commits es una de las fuentes de prueba mas fuertes para diagnosticos, implementacion y confianza tecnica.","englishUrl":"https://kyanitelabs.tech/blog/repo-archaeology-proof-assets","primaryKeyword":"repo archaeology","title":"La arqueologia de repos convierte historia en evidencia","url":"https://kyanitelabs.tech/es/blog/repo-archaeology-proof-assets"},{"description":"Lo que Kyanite agrega para que buscadores y asistentes de IA entiendan herramientas, productos, prueba y rutas de soporte.","englishUrl":"https://kyanitelabs.tech/blog/ai-discovery-llms-txt-geo","primaryKeyword":"AI discovery","title":"El descubrimiento por IA necesita mas que un sitemap","url":"https://kyanitelabs.tech/es/blog/ai-discovery-llms-txt-geo"},{"description":"Lo contrario de un DAO normal: primero la contribuci\u00f3n verificada, el token despu\u00e9s. El primer an\u00e1lisis en espa\u00f1ol del movimiento, con recibos.","primaryKeyword":"contribution economy ResonantDAO","title":"La Contribution Economy de ResonantDAO, explicada con recibos","url":"https://kyanitelabs.tech/es/blog/resonant-contribution-economy-explained"}]}