Ensō is a writing tool that helps you enter a state of flow by separating writing from editing and thus making it harder for you to edit yourself - <a href="https:&#x2F;&#x2F;enso.sonnet.io&#x2F;" rel="nofollow">https:&#x2F;&#x2F;enso.sonnet.io&#x2F;</a>After 6 years and 2 million words of daily writing I feel like I&#x27;ve learned enough to make Ensō simpler and more accessible.Related thread: <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=38025073">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=38025073</a>

Show HN: New Ensō – first public beta

the IRS recently open sourced most of Direct File, a tax tool it has been working on for a few years now. unfortunately, due to recent events, the IRS isn&#x27;t working on it anymore. I decided to pick up where they left off and I&#x27;m trying to get it ready for next tax seasonthe work behind Direct File is really interesting and I made a lot of it available online to read as well - <a href="https:&#x2F;&#x2F;docs.openfile.tax&#x2F;en&#x2F;latest&#x2F;" rel="nofollow">https:&#x2F;&#x2F;docs.openfile.tax&#x2F;en&#x2F;latest&#x2F;</a>

Show HN: A continuation of IRS Direct File that can be self-hosted

It’s available on iOS now:
<a href="https:&#x2F;&#x2F;itunes.apple.com&#x2F;app&#x2F;id6499280064" rel="nofollow">https:&#x2F;&#x2F;itunes.apple.com&#x2F;app&#x2F;id6499280064</a>We got into specialty coffee during COVID and, like many others, fell deep down the rabbit hole. Along the way, we ran into the same frustrations:- A drawer full of empty coffee bags.- No simple way to track grind size, rest dates, notes—by bean.- My coffee history scattered across photos, screenshots, notebooks, and half-memories.- The unique traits, people, and stories behind each coffee disappearing from the internet once it sold out (since coffee is an agricultural good)- In our opinion, no coffee tool really captures the flavor, emotion, and aesthetic of great coffee—from a design perspective.So we built BeanBook—a coffee notebook log beans, extract recipes, and organize your coffee life in one place with just a snap, powered by AIHere’s what it does:- Snap a bag → Auto-detects roaster, origin, process, roast date, notes, producer, farm, and more- Paste a YouTube link or photo → Extracts a structured recipe automatically- Log grind size, roast timeline, ratings &amp; notes → All saved in a clean, elegant UI- See your coffee year in review → Track habits, trends, and favorites- Ask BeanBook AI → From brew temps to bean facts, get instant answersMy co-founder and I built everything ourselves—branding, code, and UX design. If you’re into coffee (or trying to get more into it), we’d love your feedback.- Rokey &amp; Eric

Show HN: We're two coffee nerds who built an AI app to track beans and recipes

TokenDagger is a drop-in replacement for OpenAI’s Tiktoken (the tokenizer behind Llama 3, Mistral, GPT-3.*, etc.). It’s written in C++ 17 with thin Python bindings, keeps the exact same BPE vocab&#x2F;special-token rules, and focuses on raw speed.I’m teaching myself LLM internals by re-implementing the stack from first principles. Profiling TikToken’s Python&#x2F;Rust implementation showed a lot of time was spent doing regex matching. Most of my perf gains come from a) using a faster jit-compiled regex engine; and b) simplifying the algorithm to forego regex matching special tokens at all.Benchmarking code is included. Notable results show:
- 4x faster code sample tokenization on a single thread.
- 2-3x higher throughput when tested on a 1GB natural language text file.

Show HN: TokenDagger – A tokenizer faster than OpenAI's Tiktoken

What it is
A single 45 MB Windows .exe that embeds llama.cpp and a minimal Tk UI.
Copy it (plus any .gguf model) to a flash drive, double-click on any Windows PC, and you’re chatting with an LLM—no admin rights, Cloud, or network.Why I built it
Existing “local LLM” GUIs assume you can pip install, pass long CLI flags, or download GBs of extras.I wanted something my less-technical colleagues could run during a client visit by literally plugging in a USB drive.How it works
PyInstaller one-file build → bundles Python runtime, llama_cpp_python, and the UI into a single PE.On first launch, it memory-maps the .gguf; subsequent prompts stream at ~20 tok&#x2F;s on an i7-10750H with gemma-3-1b-it-Q4_K_M.gguf (0.8 GB).Tick-driven render loop keeps the UI responsive while llama.cpp crunches.A parser bold-underlines every token that originated in the prompt; Ctrl+click pops a “source viewer” to trace facts. (Helps spot hallucinations fast.)

Show HN: Local LLM Notepad – run a GPT-style model from a USB stick

International Space Station Tracker on an ESP32 CYD (Cheap Yellow Display) costing $20. Live update over Wifi, touchscreen, backlight power management. Cheap and interesting classroom STEM project. Fully open source.

Show HN: Open-Source International Space Station Tracker ESP32/Arduino for $20

Hi HN,I over-thought one too many “lol sounds fun” texts and decided to teach a model to be my wingman instead. The result is Crush Check AI —export an iMessage &#x2F; WhatsApp &#x2F; Instagram thread and get a chat report with:* crush score (0-100) based on response latency, reciprocity, sentiment shifts.* red flags like like breadcrumbing &#x2F; love-bombing* chat timeline.You can also ask questions about your conversation.Why post here
I’d love feedback on:
* whether this is something people need.
* how was the user experience.
* what features would you like to see.Thanks in advance for any roast, bug reports, or “this is useless because ___” takes. Happy to share more implementation details and happy to give away free Premium subscription in exchange for feedback!

Show HN: Crush Check – AI relationship text analyzer

Hey HN! We&#x27;re Lucas and Soheil, the founders of Praxos (<a href="https:&#x2F;&#x2F;praxos.ai">https:&#x2F;&#x2F;praxos.ai</a>). Praxos is a context manager for AI Agents, providing everything you need to build stateful agents that don&#x27;t break in production.Praxos can parse any data source, from unstructured PDFs and API streams to conversational messages, to structured databases, and transform them into a single Knowledge Graph. Everything in this graph is semantically typed and its relationships are made explicit, turning data into a clean, queryable universe of understanding that AI can use without making mistakes.Whether you need to query for the answer to a question or to extract data in a way that makes sense for the current use case, Praxos does it all, with no requerying needed. This enables AI apps to parse data end-to-end, and then act on it to deliver outputs across single-chain and multi-chain reasoning steps. Intermediate, final, and user-edited outputs can be added back to the knowledge graph, allowing Praxos to learn on the fly.When we were building in insurance, we often ran into two major problems deploying AI:First, LLMs would prove incapable of parsing documents such as property schedules and insurance policies. For reference, a property schedule may be a 50-page collection of Word, Excel, and PDF documents detailing construction, usage, and geographical information about a collection of physical properties. Recreating one object (a property) would mean combing through the files establish semantic, conceptual, spatial, and sometimes implicit linkages between the data.The outcome: relationship information would be lost, left blank, or hallucinated.Second, repeated calls to search, retrieve, and update information would sometimes lead to cascading errors. This became more frequent across complex tasks such as reading a document, fetching previous user information, performing a calculation, storing it, and then presenting it to the user.We realized that for AI to deliver more useful and accurate responses that correctly use relationships in the document, these relationships need to be made explicit. Much of the contextual information is represented without the usage of words. In turn, this means that we cannot directly interact with them programmatically, and LLMs are forced to interpret them themselves, every single time. That’s when we started building Praxos.We&#x27;ve set up a self-serve option with a free tier (up to a data cap) for hobbyists and early-adopters. For context (no pun intended), this should cover you for up to 200 document pages. You can register here: <a href="https:&#x2F;&#x2F;data.praxos.ai&#x2F;sign-up">https:&#x2F;&#x2F;data.praxos.ai&#x2F;sign-up</a>.Our first version is an SDK meant to cover you across all your data extraction, retrieval, and update needs.Here&#x27;s how it works:Organizing information: Praxos sorts information into ontologies, which are structured schemas for storing data. These allow you to introduce predefined types, attributes, and relationships that guide how the knowledge graph is built and interpreted.Processing input data: Praxos can handle any data source, ranging from PDFs to tabular data, JSONs, and dialog-like exchanges. Extraction is performed end-to-end. You don&#x27;t need to OCR, chunk, or pre-process your inputs. Processing is as simple as passing in your file and selecting an ontology.Retrieving information &#x2F; memories: For each query, Praxos searches and retrieves related stored information by leveraging a combination of graph traversal techniques, vector similarity and key-value lookups. Search objects will return both the entities&#x2F;their connections, as well as a sentence.We’d love to hear what you think! Please feel free to dive in, and share any thoughts or suggestions with us over Discord (<a href="https:&#x2F;&#x2F;discord.gg&#x2F;wVmrXD2dJA" rel="nofollow">https:&#x2F;&#x2F;discord.gg&#x2F;wVmrXD2dJA</a>).Your feedback will help shape where we take Praxos from here!

Show HN: Praxos – Context Management for AI Agents

Show HN: Private real-time dictation app for Mac

I have been working on Octelium for quite a few years now but it was open sourced only by late May 2025. Octelium, as described more in detail in the repo&#x27;s README, is simply an open source, self-hosted, unified platform for zero trust resource access that is primarily meant to be a modern alternative to corporate VPNs and remote access tools. It can operate as a remote access&#x2F;corporate VPN (i.e. alternative to Twingate, Tailscale, OpenVPN Access Server, etc...), a ZTNA&#x2F;BeyondCorp platform (i.e. alterntive to Cloudflare Access, Teleport, Google BeyondCorp, etc...), and it can also operate as an API&#x2F;AI gateway, an infrastructure for MCP and A2A architectures and meshes, an ngrok alternative, a homelab infrastructure or even as a more advanced Kubernetes ingress. It&#x27;s basically designed to operate like a unified Kubernetes-like scalable architecture for zero trust secure&#x2F;remote access that&#x27;s suitable for different human-to-workload and workload-to-workload environments. You can read more in detail the full set of main features and links about how it works in the repo&#x27;s README or directly in the docs <a href="https:&#x2F;&#x2F;octelium.com&#x2F;docs" rel="nofollow">https:&#x2F;&#x2F;octelium.com&#x2F;docs</a>

Show HN: Octelium – FOSS Alternative to Teleport, Cloudflare, Tailscale, Ngrok

I work with a team spread across Sydney, London, and SF. Last month I accidentally called my Aussie colleague at 3am their time during what I thought was a &quot;quick sync&quot;. The silence before &quot;mate... do you know what time it is here?&quot; still haunts me.Built this: <a href="https:&#x2F;&#x2F;timezig.com" rel="nofollow">https:&#x2F;&#x2F;timezig.com</a>It&#x27;s a timezone converter but it tells you if your meeting time sucks for the other person:
 - Meeting quality ratings (excellent&#x2F;good&#x2F;fair&#x2F;poor)
 - Visual indicators for day&#x2F;night
 - Shows if it&#x27;s a holiday in their country
 - Handles weird cases like Dubai&#x27;s Sunday-Thursday workweekTechnical bit: pre-generated 18k+ static pages for every city combination. Loads instantly because there&#x27;s no backend calculations. Next.js 15, no database.Still figuring out monetization (ads? affiliate links for virtual meeting tools?) but keeping it free for now.What else would make this useful? Currently tracking holidays for ~20 countries but could add more.

Show HN: Timezone converter that tells you if your meeting time sucks

I kept trying to explain how important outdoor light is and how most people aren’t getting enough. Eventually I figured that showing is more effective than telling.So we built a free app that uses your Apple Watch or iPhone to automatically track your light exposure throughout the day.It tells you if you’re getting enough, shows you how consistent you are, and rewards habits that support hormone balance.It&#x27;s in beta on TestFlight, let me know what you think!

Show HN: I built a daily sunlight tracker

Hi HN,I keep running in the same problem of each AI app “remembers” me in its own silo. ChatGPT knows my project details, Cursor forgets them, Claude starts from zero… so I end up re-explaining myself dozens of times a day across these apps.The deeper problem1. Not portable – context is vendor-locked; nothing travels across tools.2. Not relational – most memory systems store only the latest fact (“sticky notes”) with no history or provenance.3. Not yours – your AI memory is sensitive first-party data, yet you have no control over where it lives or how it’s queried.Demo video: <a href="https:&#x2F;&#x2F;youtu.be&#x2F;iANZ32dnK60" rel="nofollow">https:&#x2F;&#x2F;youtu.be&#x2F;iANZ32dnK60</a>Repo: <a href="https:&#x2F;&#x2F;github.com&#x2F;RedPlanetHQ&#x2F;core">https:&#x2F;&#x2F;github.com&#x2F;RedPlanetHQ&#x2F;core</a>What we built- CORE (Context Oriented Relational Engine): An open source, shareable knowledge graph (your memory vault) that lets any LLM (ChatGPT, Cursor, Claude, SOL, etc.) share and query the same persistent context.- Temporal + relational: Every fact gets a full version history (who, when, why), and nothing is wiped out when you change it—just timestamped and retired.- Local-first or hosted: Run it offline in Docker, or use our hosted instance. You choose which memories sync and which stay private.Why this matters- Ask “What’s our roadmap now?” and “What was it last quarter?” — timeline and authorship are always preserved.- Change a preference (e.g. “I no longer use shadcn”) — assistants see both old and new memory, so no more stale facts or embarrassing hallucinations.- Every answer is traceable: hover a fact to see who&#x2F;when&#x2F;why it got there.Try it- Hosted free tier (HN launch): <a href="https:&#x2F;&#x2F;core.heysol.ai" rel="nofollow">https:&#x2F;&#x2F;core.heysol.ai</a>- Docs: <a href="https:&#x2F;&#x2F;docs.heysol.ai&#x2F;core&#x2F;overview" rel="nofollow">https:&#x2F;&#x2F;docs.heysol.ai&#x2F;core&#x2F;overview</a>

Show HN: C.O.R.E – Opensource, user owned, shareable memory for Claude, Cursor

Audiopipe is a one-liner for denoising, diarization and transcription with demucs + pyannote + insanely-fast-whisper. 
Made it to transcribe podcasts and Dungeons And Dragons sessions, figured it could be useful.
It also has a wasm UI to load transcriptions and audio. Feel free to contribute! Feedback appreciated.

Show HN: Audiopipe – Pipeline for audio diarization, denoising and transcription

Hey HN!Pilots everywhere are required to keep a logbook of all their flying hours, aircraft, airports, and so on. Since I track everything digitally (some people still just use paper logbooks!), I put together some data visualizations and a few 3D globes to show my flying history.This globe is probably my favourite so far: <a href="https:&#x2F;&#x2F;jameshard.ing&#x2F;pilot&#x2F;globes&#x2F;all" rel="nofollow">https:&#x2F;&#x2F;jameshard.ing&#x2F;pilot&#x2F;globes&#x2F;all</a>If you’ve got ideas for other graphs or ways to show this kind of data, I’d love to hear them!

Show HN: I'm an airline pilot – I built interactive graphs/globes of my flights

We’ve been building local and on-prem agent workflows for open-source LLMs. Engines like Ollama or vLLM ship with no auth, so every team ends up writing the same JWT proxy.Attach Gateway is a single process that sits in front of any model server and handles the boring bits:- verifies OIDC &#x2F; DID JWTs 
- adds X-Attach-User and X-Attach-Session headers so downstream agents share the same identity 
- optional &#x2F;a2a&#x2F;tasks&#x2F;send endpoint for Google-style A2A hand-offs 
- mirrors prompts + completions to Weaviate (runs in Docker)One `pip install attach-dev`, export a token, run `attach-gateway`, and your local Ollama is protected in under 60 seconds.Repo: <a href="https:&#x2F;&#x2F;github.com&#x2F;attach-dev&#x2F;attach-gateway">https:&#x2F;&#x2F;github.com&#x2F;attach-dev&#x2F;attach-gateway</a> 
PyPI: <a href="https:&#x2F;&#x2F;pypi.org&#x2F;project&#x2F;attach-dev&#x2F;" rel="nofollow">https:&#x2F;&#x2F;pypi.org&#x2F;project&#x2F;attach-dev&#x2F;</a>Would love feedback – especially from teams doing multi-agent or on-prem work.

Show HN: Attach Gateway – one-command OIDC/DID auth for local LLMs

Show HN: Summle – A little maths Game

Show HN: ArcFont – Font Embedding Model

QuizKnit is an open source quiz creator to create quizzes by pasting in text or selecting from a large range of historical categories.Started as a very simple app for me to play around with OpenAI’s API last year then morphed into a portfolio project during my job search earlier this year. Now happily employed so decided to open source it.Right now, a user can create a quiz, take a quiz, save it and share the quiz with other people using a URL.You can try out the full working application at <a href="https:&#x2F;&#x2F;quizknit.com" rel="nofollow">https:&#x2F;&#x2F;quizknit.com</a>Github Links: Frontend: <a href="https:&#x2F;&#x2F;github.com&#x2F;jibolash&#x2F;quizknit-react">https:&#x2F;&#x2F;github.com&#x2F;jibolash&#x2F;quizknit-react</a> , Backend: <a href="https:&#x2F;&#x2F;github.com&#x2F;jibolash&#x2F;quizknit-api">https:&#x2F;&#x2F;github.com&#x2F;jibolash&#x2F;quizknit-api</a>

Show HN: QuizKnit, an open source quiz creator

I built a simple but effective Sharpe Ratio calculator that gives the full historical variation of it.
Should I add other rations like Calmar and Sortino?

Show HN: Sharpe Ratio Calculation Tool

I spent the past 2 weeks making this toy programming language.I basically forked the Go scanner&#x2F;parser, and changed the syntax to have functions return a single value. This enable proper Result&#x2F;Option type to be used as well as propagating errors with an operator.I also wanted to have short &quot;type inferred&quot; anonymous functions to be able to use functions like Map&#x2F;Reduce&#x2F;Filter, without having to use 100 characters to specify the types.

Show HN: AGL a toy language that compiles to Go

Show HN: Vet – A tool for safely running remote shell scripts

Hi HN! I built MailMap (<a href="https:&#x2F;&#x2F;mailmap.site" rel="nofollow">https:&#x2F;&#x2F;mailmap.site</a>) – a platform where you can create interactive map stories by simply sending an email.## How it works:
1. Take a photo with GPS data
2. Email it to submit@mailmap.site with your story
3. Your story appears on an interactive map within seconds
4. Get a unique shareable URLNo app downloads, no registrations required for basic use. Just email → map story.## Technical details:
- Automatically extracts GPS coordinates from image EXIF data
- Alternative: registered users can add stories directly on the map (click location + wait 3 seconds)
- Real-time processing and map updates
- Built for the Postmark Challenge: Inbox Innovators
- Supports common image formats (JPEG, PNG, GIF)
- Community moderation system
- Mobile-responsive Google Maps integration## Why email-first?
I wanted to remove friction from sharing location-based stories. Everyone knows how to send an email, but not everyone wants to download another app or create another account. Email is universal.The platform also supports registered users who can manually add&#x2F;edit stories directly on the map, but the core experience works entirely through email.## Current features:
- Instant story publishing via email
- Interactive map exploration
- Story voting and sharing
- Community moderation
- Responsive design
- Direct map editing for registered usersTry it out: send a geotagged photo to submit@mailmap.site or explore existing stories at <a href="https:&#x2F;&#x2F;mailmap.site" rel="nofollow">https:&#x2F;&#x2F;mailmap.site</a>What do you think? Any feedback on the concept or technical implementation?---*Demo:* <a href="https:&#x2F;&#x2F;mailmap.site" rel="nofollow">https:&#x2F;&#x2F;mailmap.site</a> 
*Email:* submit@mailmap.site 
*Help:* <a href="https:&#x2F;&#x2F;mailmap.site&#x2F;help" rel="nofollow">https:&#x2F;&#x2F;mailmap.site&#x2F;help</a>

Show HN: MailMap – Turn emails into interactive stories on Google Maps

i made sink. it&#x27;s a simple little tool that continuously syncs folders between 2 devices. no cloud, no email, flash drives, no bs.it just uses your local wifi. run it on your machines, tell them to trust each other, and you&#x27;re set. and if you manage to edit the same file at once, it handles the conflict and saves both copies.for anyone who just wants to get files from point a to b without the headache. hope it makes your life a bit less annoying.github: <a href="https:&#x2F;&#x2F;github.com&#x2F;sirbread&#x2F;sink">https:&#x2F;&#x2F;github.com&#x2F;sirbread&#x2F;sink</a>
binary: <a href="https:&#x2F;&#x2F;github.com&#x2F;sirbread&#x2F;sink&#x2F;releases&#x2F;tag&#x2F;v0.1">https:&#x2F;&#x2F;github.com&#x2F;sirbread&#x2F;sink&#x2F;releases&#x2F;tag&#x2F;v0.1</a>

Show HN: Sink – Sync any directory with any device on your local network

The idea is simple: a Python dictionary that uses semantic similarity during lookup. Say you&#x27;re working on user preference extraction for an AI agent. In some cases, the model may output something like:```
{
 &quot;preference&quot;: &quot;Database&quot;,
 &quot;value&quot;: &quot;sqlite3&quot;
}
```but it (or a different model) may later request the user preference using a subtly different key:```
{
 &quot;action&quot;: &quot;get_user_preference&quot;,
 &quot;preference&quot;: &quot;Database System&quot;
}
```If we parse these responses and store them in a standard dictionary, we&#x27;ll miss the similarity between the two preference keys &quot;Database&quot; and &quot;Database System&quot;, whereas with SemanticDictionary, an embedding model would calculate the distance between the embedding vectors corresponding to the the preference keys prior to lookup.I figured others may have similar needs so I thought I&#x27;d package it up for distribution. Hope you find something cool to do with this!

Show HN: Semantic-dictionary – A Python dictionary with semantic lookup

I recently built a small open-source tool to benchmark different LLM API endpoints — including OpenAI, Claude, and self-hosted models (like llama.cpp).It runs a configurable number of test requests and reports two key metrics:
 • First-token latency (ms): How long it takes for the first token to appear
 • Output speed (tokens&#x2F;sec): Overall output fluencyDemo: <a href="https:&#x2F;&#x2F;llmapitest.com&#x2F;" rel="nofollow">https:&#x2F;&#x2F;llmapitest.com&#x2F;</a>
Code: <a href="https:&#x2F;&#x2F;github.com&#x2F;qjr87&#x2F;llm-api-test">https:&#x2F;&#x2F;github.com&#x2F;qjr87&#x2F;llm-api-test</a>The goal is to provide a simple, visual, and reproducible way to evaluate performance across different LLM providers, including the growing number of third-party “proxy” or “cheap LLM API” services.It supports:
 • OpenAI-compatible APIs (official + proxies)
 • Claude (via Anthropic)
 • Local endpoints (custom&#x2F;self-hosted)You can also self-host it with docker-compose.
Config is clean, adding a new provider only requires a simple plugin-style addition.Would love feedback, PRs, or even test reports from APIs you’re using. Especially interested in how some lesser-known services compare.

Show HN: A tool to benchmark LLM APIs (OpenAI, Claude, local/self-hosted)

I&#x27;ve been working for a good 9 months now writing a .net to C transpiler (<a href="https:&#x2F;&#x2F;github.com&#x2F;KallDrexx&#x2F;dntc">https:&#x2F;&#x2F;github.com&#x2F;KallDrexx&#x2F;dntc</a>) as a &quot;I wonder if I can&quot; type of project (with the intention of eventually being able to do embedded development in C# instead of C).Along the way I was trying to find good milestones to test the transpiler with. As I learned about Linux eBPF I realized that ended up being an interesting use case and created the DotnetEbpf prototype.It allows you to write Linux kernel side eBPF applications fully in C#. While with some of the limitations it may seem like it&#x27;s limiting the usefulness of C#, it does allow for better syntax than C and as well as unit testing capabilities (since each native eBPF function can be populated with C# native code that is ignored when transpiled, in addition to generic capabilities).In the future I&#x27;d like to flesh out the ability for correct wiring up of data structures between both the C# kernel side and a C# user space side (and the ability to install an eBPF kernel application from a C# user space application), which would enable writing full stack eBPF security software fully in C#.

Show HN: DotnetEbpf - Write Linux eBPF kernel applications in C#

Show HN: Ravana, Multi AI Assistant Browser Desktop App

I have just completed the MySQL parser that will be soon used in MIST.The story behind is that TiDB MySQL parser, which is by far the most MySQL compatible parser, was depending on some types called &quot;parser_driver&quot;.Parser Driver contains types from TiDB, but unfortunately, TiDB is not going to moving MySQL types, which is heavy entangled with TiDB, out as a independent Go package. GitHub issue:16381To me, TiDB itself is mixing all its open-sourced modules as a very opinionated monolith.I was porting MIST to WASM binary and TiDB syscalls will be stupidly included, compiled, and making static linking failed when targeting WASM.Thus I spent a few days to extract the existing parser package and parser driver and re-export the entangled TiDB types into an ACTUALLY WORKING MySQL parser go package, which is truly cross-platform and ready to port to WASM.Repository: <a href="https:&#x2F;&#x2F;github.com&#x2F;abbychau&#x2F;mysql-parser">https:&#x2F;&#x2F;github.com&#x2F;abbychau&#x2F;mysql-parser</a>
Interactive Parser Demo: <a href="https:&#x2F;&#x2F;abbychau.github.io&#x2F;mysql-parser&#x2F;" rel="nofollow">https:&#x2F;&#x2F;abbychau.github.io&#x2F;mysql-parser&#x2F;</a>

Show HN: My Cross-Platform MySQL Parser

Hello!I&#x27;m Andrew Arrow, a developer and Final Cut Pro user. My history with FCP goes back years and years, I tried to be a video editor as a career but ended up a software developer. I purchased the full version of the software a long, long time ago. And to Apple&#x27;s credit, every single time I goto the App Store on any mac I still have access to download it. I must have been through 20, 30? I don&#x27;t know, but a lot of macs over the years. Just recently I downloaded it again and noticed the XML Export feature.Wow. There&#x27;s a lot of stuff there. I mean a lot. I started playing around with generating the XML and that has lead to this project.<a href="https:&#x2F;&#x2F;github.com&#x2F;andrewarrow&#x2F;cutlass&#x2F;blob&#x2F;main&#x2F;README.md">https:&#x2F;&#x2F;github.com&#x2F;andrewarrow&#x2F;cutlass&#x2F;blob&#x2F;main&#x2F;README.md</a>Read through those go structs and xml tags. Keep going. And going. It just never stops! It&#x27;s a very sophisticated XML format that can describe any timeline, effect, or animation you can imagine. Apple&#x27;s documentation is scattered, the XML is notoriously finicky, and one wrong attribute crashes your import.And of course with Claude Code (or others, pick your favorite AI) you can have your AI write some amazing things in go using cutlass.```go
&#x2F;&#x2F; Generate a timeline with precise keyframe animationsvideo := fcp.Video{<pre><code> Ref: assetID,

 Offset: &quot;0s&quot;, 

 Duration: fcp.ConvertSecondsToFCPDuration(10.0),

 AdjustTransform: &amp;fcp.AdjustTransform{

 Params: []fcp.Param{{

 Name: &quot;position&quot;,

 KeyframeAnimation: &amp;fcp.KeyframeAnimation{

 Keyframes: []fcp.Keyframe{

 {Time: &quot;0s&quot;, Value: &quot;0 0&quot;},

 {Time: &quot;240240&#x2F;24000s&quot;, Value: &quot;100 50&quot;},

 },

 },

 }},

 },</code></pre>
}
```So I haven&#x27;t been writing code like that with my human brain. I spend most of my time just telling Claude the video I want to create, and he uses the cutlass library to achieve it. I&#x27;m hooking up FCP directly to Claude. A different kind of AI Video generation.Would love feedback, thanks for reading this far!
-aa