I’ve been building a browser IDE called *Pyxis CodeCanvas*, designed mainly for iPad and quick coding sessions.It’s still a work in progress (expect some bugs!), but I’d love to get feedback — especially from devs interested in browser runtimes or local-first tools.Pyxis aims to be a “1-second-to-open” IDE that runs entirely client-side — no backend, no cloud. 
It uses OPFS + IndexedDB for persistent storage and runs smoothly even on Safari.Currently supports:
- TypeScript &#x2F; JavaScript &#x2F; Python 
- Partial Node.js runtime 
- Git&#x2F;GitHub integration (push, pull, clone, even private repos)It’s optimized for low-memory devices, supports GitHub Pages deploys directly, and includes AI-assisted code review and markdown preview.I sometimes use Pyxis itself to work on Pyxis (not fully, but enough to be practical). 
The goal is to fill the gap between full VS Code and just editing text in a browser tab.*Repo:* <a href="https:&#x2F;&#x2F;github.com&#x2F;Stasshe&#x2F;Pyxis-CodeCanvas" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;Stasshe&#x2F;Pyxis-CodeCanvas</a> 
*Demo:* <a href="https:&#x2F;&#x2F;stasshe.github.io&#x2F;Pyxis-CodeCanvas&#x2F;" rel="nofollow">https:&#x2F;&#x2F;stasshe.github.io&#x2F;Pyxis-CodeCanvas&#x2F;</a>Would love to hear what you think — what’s broken, missing, or surprising!

Show HN: Pyxis CodeCanvas a lightweight, client-side IDE for iPad and browsers

Show HN: MacOS Live Screensaver – A screensaver that plays live video streams

My wife loves it - she&#x27;s the one I made it for. She also helped shape it by playing early versions and giving lots of feedback. I hope you&#x27;ll like it, too.Here are some of the technical details for those who enjoy reading about that. The client was built in C++ using Axmol engine, while the server was built in Go. They communicate via HTTP and Websockets. The server is hosted on an AWS EC2 m6g.large instance, together with nginx, PostgreSQL and admin in Laravel. They are each running in their own Docker container, but all on the same instance. The game leverages Firebase Auth for user authentication. The code is hosted on Github and uses Github actions to compile, dockerize and publish a new version of any server component to the cloud. Is it overcomplicated for such a small game? Possibly yes, but I don&#x27;t mind. It was also a learning experience for me to set it all up by myself. The total cost of the cloud is around US $30 per month, and the game doesn&#x27;t earn enough to cover for it, but it doesn&#x27;t matter, as I would&#x27;ve had an instance running there regardless (I have a 3-year commitment).If you have any questions, just ask away. And if you want to try it for yourself, here&#x27;s the link:<a href="https:&#x2F;&#x2F;www.anagramarena.com&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.anagramarena.com&#x2F;</a>

Show HN: I made an anagram word game for mobile in C++ and Go

Built a developer-friendly BaaS around PostgreSQL + PostGIS. Instant APIs, real-time updates, self-hostable Docker image. Feedback welcome

Show HN: Centia.io – Open PostgreSQL/PostGIS back end for developers

TLDR: I’m expanding the family of text-splitting Chonky models with new multilingual model.You can learn more about this neural approach in a previous post: <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=43652968">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=43652968</a>Since the release of the first distilbert-based model I’ve released two more models based on a ModernBERT. All these models were pre-trained and fine-tuned primary on English texts.But recently mmBERT(<a href="https:&#x2F;&#x2F;huggingface.co&#x2F;blog&#x2F;mmbert" rel="nofollow">https:&#x2F;&#x2F;huggingface.co&#x2F;blog&#x2F;mmbert</a>) has been released. This model pre-trained on massive dataset that contains 1833 languages. So I had an idea of fine-tuning a new multilingual Chonky model.I’ve expanded training dataset (that previously contained bookcorpus and minipile datasets) with Project Gutenberg dataset which contains books in some widespread languages.To make the model more robust for real-world data I’ve removed punctuation for last word for every training chunk with probability of 0.15 (no ablation was made for this technique though).The hard part is evaluation. The real-world data are typically OCR&#x27;ed markdown, transcripts of calls, meeting notes etc. and not a clean book paragraphs. I didn’t find such labeled datasets. So I used what I had: already mentioned bookcorpus and Project Gutenberg validation, Paul Graham essays, concatenated 20_newsgroups.I also tried to fine-tune the bigger mmBERT model (mmbert-base) but unfortunately it didn’t go well — metrics are weirdly lower in comparison with a small model.Please give it a try. I&#x27;ll appreciate a feedback.The new multilingual model: <a href="https:&#x2F;&#x2F;huggingface.co&#x2F;mirth&#x2F;chonky_mmbert_small_multilingual_1" rel="nofollow">https:&#x2F;&#x2F;huggingface.co&#x2F;mirth&#x2F;chonky_mmbert_small_multilingua...</a>All the Chonky models: <a href="https:&#x2F;&#x2F;huggingface.co&#x2F;mirth" rel="nofollow">https:&#x2F;&#x2F;huggingface.co&#x2F;mirth</a>Chonky wrapper library: <a href="https:&#x2F;&#x2F;github.com&#x2F;mirth&#x2F;chonky" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;mirth&#x2F;chonky</a>

Show HN: Chonky – a neural text semantic chunking goes multilingual

Hey HNI built ImageConverter.dev
 because I got tired of “free” image converter sites that force uploads, or throttle conversions.So I made a tool that runs 100% client-side — meaning your images never leave your device. It’s built for speed, simplicity, and privacy.What it doesConvert images between JPG, PNG, WebPWorks offline once loaded (PWA support)How it worksIt uses the Canvas API and WebAssembly to handle conversions directly in the browser.
There’s no upload, no tracking, no server costs, and it’s fast even on mid-range devices.Why I built itI wanted an instant, no-ads, privacy-safe way to handle images for my personal projects — something lightweight enough to replace desktop tools.Try it<a href="https:&#x2F;&#x2F;imageconverter.dev" rel="nofollow">https:&#x2F;&#x2F;imageconverter.dev</a>Would love feedback from the community — especially on:Performance on different browsers&#x2F;devicesThanks for reading!

Show HN: A fast, privacy-first image converter that runs in browser

Hi HN, we’re Jamie and Matti, co-founders of Twigg.During our master’s we continually found the same pain points cropping up when using LLMs. The linear nature of typical LLMs interfaces - like ChatGPT and Claude - made it really easy to get lost without any easy way to visualise or navigate your project.Worst of all, none of them are well suited for long term projects. We found ourselves spending days using the same chat, only for it to eventually break. Transferring context from one chat to another is also cumbersome. We decided to build something more intuitive to the ways humans think.We started with two simple ideas. Enabling chat branching for exploring tangents, and an interactive tree diagram to allow for easy visualisation and navigation of your project.Twigg has developed into an interface for context management - like “Git for LLMs”. We believe the input to a model - or the context - is fundamental to its performance. To extract the maximum potential of an LLM, we believe the users need complete control over exactly what context is provided to the model, which you can do using simple features like cut, copy and delete to manipulate your tree.Through Twigg, you can access a variety of LLMs from all the major providers, like ChatGPT, Gemini, Claude, and Grok. Aside from a standard tiered subscription model (free, plus, pro), we also offer a Bring Your Own Key (BYOK) service, where you can plug and play with your own API keys.Our target audience are technical users who use LLMs for large projects on a regular basis. If this sounds like you, please try out Twigg, you can sign up for free at <a href="https:&#x2F;&#x2F;twigg.ai&#x2F;" rel="nofollow">https:&#x2F;&#x2F;twigg.ai&#x2F;</a>. We would love to get your feedback!

Show HN: Git for LLMs – A context management interface

Hi HN! I would like to present my project called TOMMY, which turns ESP32 devices into motion sensors that work through walls and obstacles using Wi-Fi sensing.TOMMY started as a project for my own use. I was frustrated with motion sensors that didn&#x27;t detect stationary presence and left dead zones everywhere. Presence sensors existed but were expensive and needed one per room. I explored echo localization first, but microphones listening 24&#x2F;7 felt too creepy. Then I discovered Wi-Fi sensing - a huge research topic but nothing production-ready yet. It ticked all the boxes: could theoretically detect stationary presence through breathing&#x2F;micromovements and worked through walls and furniture so devices could be hidden away.Two years and dozens of research papers later, TOMMY has evolved into software I&#x27;m honestly quite proud of. Although it doesn&#x27;t have stationary presence detection yet (coming Q1 2026) it detects motion really well. It works as a Home Assistant Add-on or Docker container, supports a range of ESP32 devices, and can be flashed through the built-in tool or used alongside existing ESPHome setups.I released the first version a couple of months ago on Home Assistant&#x27;s subreddit and got a lot of interest and positive feedback. More than 200 people joined the Discord community and almost 2,000 downloaded it.Right now TOMMY is in beta, which is completely free for everyone to use. I&#x27;m also offering free lifetime licenses to every beta user who joins the Discord channel.You can read more about the project on <a href="https:&#x2F;&#x2F;www.tommysense.com" rel="nofollow">https:&#x2F;&#x2F;www.tommysense.com</a>. Please join the Discord channel if you are interested in the project.A note on open source: There&#x27;s been a lot of interest in having TOMMY as an open source project, which I fully understand. I&#x27;m reluctant to open source before reaching sustainability, as I&#x27;d love to work on this full time. However, privacy is verifiable - it&#x27;s 100% local with no data collection (easily confirmed via packet sniffing or network isolation). Happy to help anyone verify this.

Show HN: Tommy – Turn ESP32 devices into through-wall motion sensors

Hi HN!We got frustrated with the fragmented experience of exploring &amp; creating across our file manager, the web and document apps. Lots of manual searching, opening windows &amp; tabs, scrolling, and ultimately copying &amp; pasting into a document editor.Surf is a desktop app meant for simultaneous research and thinking to minimize the grunt work. It’s made of two parts:1) A multi-media library where you can save and organize files and webpages into collections called Notebooks.2) A LLM-powered smart document which you can auto-generate using the context from any stored page, tab or entire notebook. This document contains deep links back to the source material — like a page of a PDF or timestamp in a YouTube video. Unlike Deep Research products (or NotebookLMs chat) the entire thing is editable. The user also stays in the loop.With a technology like AI, context &#x2F; data is proving to be king. We think it should stay under the user’s control, with minimal lock in: where you can own &amp; export, and plug &amp; play with different models. That’s why Surf is:- Open Source on GitHub
- Open (&amp; Local Data): the data saved in Surf is stored on your local machine in open and accessible formats and mostly works offline.
- Open Model Choice: you can choose which models you use with Surf, and can add custom &amp; Local LLMsEarly users include students &amp; researchers who are learning and doing thematic research using Surf.Github repo: <a href="https:&#x2F;&#x2F;github.com&#x2F;deta&#x2F;surf&#x2F;" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;deta&#x2F;surf&#x2F;</a>Website: <a href="https:&#x2F;&#x2F;deta.surf&#x2F;" rel="nofollow">https:&#x2F;&#x2F;deta.surf&#x2F;</a>

Show HN: Deta Surf – An open source and local-first AI notebook

Show HN: Building a good habit? Join us:)

Hello, this is my first project!I used nothing but free web tools. And it seems to work.I couldn&#x27;t find a simple portfolio performance tool on the web, where I could just paste cash inflows, outflows and current value of my portfolio and get the portfolio performance calculated. Everything I found was in some way cumbersome and cluttered with ads.I used Perplexity for the coding and Codeberg for the hosting.I verified the output manually for some examples and all seems fine. Maybe it is useful for others too. If anyone can spot a problem in this small project, I would be very grateful to hear about it.

Show HN: I made a simple portfolio performance calculator and it seems to work

I’ve been a long-time Snowplow user and unofficial evangelizer. I have deep respect for its founders, Alex and Yali, who I met a few times.What made me fall in love with Snowplow was that it was unopinionated, gave access to raw event data, and was truly open source. Back in 2013, that changed everything for me. I couldn’t look at GA the same way again.Over the years, analytics moved into SQL warehouses driven by cheaper CPU&#x2F;storage, dbt, reproducibility, and transparency. I saw the need for a democratized Snowplow pipeline and launched a hosted version in 2019.In January 2024, Snowplow changed its license (SLULA), effectively ending open-source Snowplow by restricting production use. When that happened, I realized the spirit of open data and open architecture was gone.A week later, I forked it, I wanted to keep the idea alive.OpenSnowcat keeps the original collector and enricher under Apache 2.0 and stays fully compatible with existing Snowplow pipelines. We maintain it with regular patches, performance optimizations, and integrations with modern tools like Warpstream Bento for event processing&#x2F;routing.The goal is simple: keep open analytics open.Would love to hear how others in the community think we can preserve openness in data infrastructure as “open source” becomes increasingly commercialized.That&#x27;s it, I should have posted here earlier but now felt right.

Show HN: OpenSnowcat – A fork of Snowplow to keep open analytics alive

a fast zero-copy sqlite implementation only for dumping the contents of a sqlite3 file

Show HN: Sqlite3-dump - a fast SQLite to CSV and parquet

We built Nostr Web, a new way to publish and host websites that live entirely on the Nostr network instead of centralized servers.Each website is a collection of signed, verifiable Nostr events distributed across relays—so it can’t be taken down, censored, or lost.It includes:
• DNS TXT records for domain-based discovery (_nweb.domain.com)
• CLI publisher tool for versioned deployments (nw-publisher)
• Browser extension (nw-extension) for native browsing experience
• Relay v1.3.5 support for Nostr Web event kindsTry the live demo: <a href="https:&#x2F;&#x2F;nweb.shugur.com" rel="nofollow">https:&#x2F;&#x2F;nweb.shugur.com</a>Repos: <a href="https:&#x2F;&#x2F;github.com&#x2F;Shugur-Network&#x2F;relay" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;Shugur-Network&#x2F;relay</a> | <a href="https:&#x2F;&#x2F;github.com&#x2F;Shugur-Network&#x2F;nw-nips" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;Shugur-Network&#x2F;nw-nips</a> | <a href="https:&#x2F;&#x2F;github.com&#x2F;Shugur-Network&#x2F;nw-publisher" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;Shugur-Network&#x2F;nw-publisher</a> | <a href="https:&#x2F;&#x2F;github.com&#x2F;Shugur-Network&#x2F;nw-extention" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;Shugur-Network&#x2F;nw-extention</a>Would love feedback from the HN community—on protocol design, relay performance, or UX ideas for improving decentralized web publishing.

Show HN: Nostr Web – decentralized website hosting on Nostr

Hi HN ,Six months ago, I built JustGrind, an AI habit coach. Coding it was fun; shipping it? Nightmare. Marketing it? Even worse.I realized the hardest part of building a product isn’t the coding—it’s getting people to actually see it. So I built Postnix AI, a tool that helps indie hackers, creators, and solo founders craft posts that get attention:- Fetches top-performing posts in your target communities- Analyzes what makes them work- Helps you write your own posts that feel natural and hit the right vibe- Suggests flairs, formatting, and style tweaks that increase visibility and engagementKey lesson: no matter how good your product is, if nobody knows it exists, it doesn’t matter.It’s still early days, but people are already getting traction with their posts. Would love feedback from the HN community!Demo: <a href="https:&#x2F;&#x2F;www.tiktok.com&#x2F;@postnixai&#x2F;video&#x2F;7563596577915391239" rel="nofollow">https:&#x2F;&#x2F;www.tiktok.com&#x2F;@postnixai&#x2F;video&#x2F;7563596577915391239</a>

Show HN: Postnix AI – Helps Indie Hackers Get Their Work Seen

&gt; &quot;In a world without nil safety, one gem dares to ask: &#x27;What if we just guessed?&#x27;&quot;

Show HN: LLM Rescuer – Fixing the billion dollar mistake in Ruby

An honest to god, non-algorithmic reverse chrono list of tech news that passes my signal-to-noise tests, updated hourly.A lightweight a page design as I&#x27;ve been able to keep; simple, clean, fast. No commercial features or aspirations - this is a passion project, something I&#x27;ve been fooling around with on and off for decades.There&#x27;s a &quot;Top&quot; view too with an LLM edited front page &amp; summary, and categorized views for a large number of topics - see the Directory. A few more buried features to explore, but the fundamental use case is pop in, scan, exit - fast and concise.Your feedback would be appreciated!

Show HN: I built a tech news aggregator that works the way my brain does

Hi HN, I&#x27;m Nick and I built Inspec with my partner Caitlin (she&#x27;s the interior designer, I&#x27;m the developer).Inspec lets interior designers create and manage FF&amp;E schedules - basically spec sheets listing everything that goes in a room before the furniture: flooring, paint, countertops, lighting, fixtures, etc.This process is currently exclusively done using Excel, and while this works to some degree, it comes with a lot of manual work and clunkiness. When Caitlin started working remotely, I watched her deal with these issues daily. So I spent the past 9 months building a better solution.Key Features:- Realtime collaboration- Revision control (versioning)- Professional PDF exports- QR codes to print out on-site for contractors, builders etc. to always have the latest version- Completely customisable fields and a familiar workflow to Excel to keep learning curve smallWhen we first started building this we thought it was completely unique, but turns out there are 2 competitors. I think our positioning is quite different though as we seek to only replace this document creation, rather than overhaul the entire project management and try to be a complete solution. So that’s helped to guide our focus, but also validated that interior designers are looking to modernise their software.Tech stack:I used my SaaS boilerplate which is just a modified T3 stack template with everything set up just how I like it. So Next.js with the pages router, TS, Trpc, Prisma + PostgreSQL, and TailwindCSS. I’ve also got Pusher for realtime collaboration, Redis and BullMQ for keeping a seperate background worker for PDF and web scraping jobs.I can’t imagine many interior designers are on HN, but figured I would share this for anyone who wants to take a look :) feedback always welcome! The landing page is at <a href="https:&#x2F;&#x2F;inspec.design" rel="nofollow">https:&#x2F;&#x2F;inspec.design</a>

Show HN: Inspec – Specification scheduling software for interior designers

Show HN: Cuq – Formal Verification of Rust GPU Kernels

I&#x27;ve built this to make it easy to host your own infra for lightweight VMs at large scale.Intended for exec of AI-generated code, for CICD runners, or for off-chain AI DApps. Mainly to avoid Docker-in-Docker dangers and mess.Super easy to use with CLI &#x2F; Python SDK, friendly to AI engs who usually don&#x27;t like to mess with VM orchestration and networking too much.Defense-in-depth philosophy.Would love to get feedback (and contributors: clear &amp; exciting roadmap!), thx

Show HN: Katakate – Dozens of VMs per node for safe code exec

I built it for myself because ngrok didn&#x27;t let me add one more domain unless I paid $12 more. I probably should&#x27;ve looked for alternatives before building my own, but my grudge got in the way . Once I started, I wanted to make it better than the other options. Silly, I know. No one probably cares.I plan to open source it sometime next year because I’ve got other projects to finish first.
It&#x27;s free until I deplete my cloud credits, then it will switch to a tier-based model with a free option.

Show HN: Wsgrok – one of many ngrok alternatives

Hey HN,My name is Brett and I&#x27;m excited to share ScreenAsk with you today!ScreenAsk makes it easy to collect screen recordings from your customers by simply sending a link.At my SaaS company, we spend hours every week teaching customers to record their screen to show us support issues.They have to sign up for a new tool, download and install it, create the recording, and then upload it somewhere to send it over…Most of the time spent on support wasn&#x27;t fixing the issue but rather understanding it.I built ScreenAsk to solve this exact problem, making it simple to see what your customers see:- Send over your recording link- They follow a few easy steps to record their screen- You instantly see what they’re experiencingNo sign up, no installing extra software, and no uploading to another service just to share.You can get notified via Email + Slack + Zapier + Webhooks when somebody records, and recordings include transcription + AI summaries for quick scanning.We also offer a widget that can be embedded in your site and is fully customizable + controllable with javascript.- Show &#x2F; hide it when you want- Change colors and language- Listen for a recording and populate a form field with the viewing link- Add metadata like name, email, ID to the recording- Capture network and consoleI’d be grateful if you gave it a spin! You get 10 free recordings per month and a personalized recording link: <a href="https:&#x2F;&#x2F;screenask.com" rel="nofollow">https:&#x2F;&#x2F;screenask.com</a>Launch tweet + demos + discussion: <a href="https:&#x2F;&#x2F;x.com&#x2F;balindenberg&#x2F;status&#x2F;1981394673173205418" rel="nofollow">https:&#x2F;&#x2F;x.com&#x2F;balindenberg&#x2F;status&#x2F;1981394673173205418</a>

Show HN: ScreenAsk – Free Screen Recording Links for Customer Support

After Pocket shut down, I started working on a new bookmarking solution. My main goal was to make bookmarks searchable, something that Pocket was not good at.So I built Pensive, a bookmarking solution that saves the full page content and makes it searchable with full-text search. You can add pages using a browser extension or a Telegram bot (for saving on mobile).It is written in Go with PostgreSQL, all Dockerized on a $5 Hetzner server, and the front end is built with HTMX.I have also added embeddings (using Gemini Flash Lite) to let LLMs interact with your bookmarks contextually.It is stable enough that I now use it daily. I’m considering publishing it as open source, but first I want to have a proper version ready. Feel free to email me if you’re interested in helping out or if you have prior experience with open source.Feel free to try it at <a href="https:&#x2F;&#x2F;getpensive.com" rel="nofollow">https:&#x2F;&#x2F;getpensive.com</a>

Show HN: Pensive – A bookmarking tool with full-text search and LLMs

I got tired of playwright-mcp eating through Claude&#x27;s 200K token limit, so I built this using the new Claude Skills system. Built it with Claude Code itself.Instead of sending accessibility tree snapshots on every action, Claude just writes Playwright code and runs it. You get back screenshots and console output. That&#x27;s it.314 lines of instructions vs a persistent MCP server. Full API docs only load if Claude needs them.Same browser automation, way less overhead. Works as a Claude Code plugin or manual install.Token limit issue: <a href="https:&#x2F;&#x2F;github.com&#x2F;microsoft&#x2F;playwright-mcp&#x2F;issues&#x2F;889" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;microsoft&#x2F;playwright-mcp&#x2F;issues&#x2F;889</a>Claude Skills docs: <a href="https:&#x2F;&#x2F;docs.claude.com&#x2F;en&#x2F;docs&#x2F;claude-code&#x2F;skills" rel="nofollow">https:&#x2F;&#x2F;docs.claude.com&#x2F;en&#x2F;docs&#x2F;claude-code&#x2F;skills</a>

Show HN: Playwright Skill for Claude Code – Less context than playwright-MCP

hey hn!I (re)built this TUI tool for browsing BBC News in the terminal, it uses an RSS feed for getting headlines and previews and you can read articles too.Try it out and let me know what you think! :)

Show HN: bbcli – A TUI and CLI to browse BBC News like a hacker

Hello world! I would like to share that I have created a simple open-source Python CLI app to check what&#x27;s hogging all the disk space! You can install it with pip. It&#x27;s like space but Zpace.pip install zpaceand find the big files consuming your disk space. Be it apps, virtual environments, machine learning models etc. Just runzpaceOnce you find one, that you can say &quot;bye&quot; to, just runrm -rf &#x2F;i_dont_need&#x2F;this&#x2F;file&#x2F;right_nowto get rid of it.It was born out of frustration while lack of disk space prevented me to use my laptop properly. It has been very useful to me so far. I hope that it can be useful to you as well. Feel free to check it out.Currently tested on MacOS only

Show HN: Check what is hogging your disk zpace

Show HN: Databroom – Clean data with GUI or CLI and generate Python/R script

Show HN: Integrating semantic commits directly into Git could unlock a new level

This is a recent addition to Vexlio which I think the HN crowd may find interesting or useful.TL;DR: easy creation of interactive diagrams, meaning diagrams that have mouse click&#x2F;hover hooks that you can use to display pop-up content. The end result can be shared with a no-sign-in-required web link.My thought is that this is useful for system docs, onboarding or user guides, presentations, etc. Anything where there is a high-level view that should remain uncluttered + important metadata or details that still need to be available somewhere.You can try it out without signing up for anything, just launch the app here (<a href="https:&#x2F;&#x2F;app.vexlio.com&#x2F;" rel="nofollow">https:&#x2F;&#x2F;app.vexlio.com&#x2F;</a>), create a shape, select it with the main pointer tool and then click &quot;Add popup&quot; on the context toolbar.I&#x27;d be grateful for any and all feedback!

Show HN: Create interactive diagrams with pop-up content

I was using LLM frameworks everywhere but had no idea what was happening inside them. One day I needed to optimize something and realized I couldn&#x27;t. Hard truth: I didn&#x27;t understand the fundamentals, just which framework function to call.So I stripped everything away. No abstractions. Just Python, HTTP requests, and the OpenAI&#x2F;Anthropic APIs.What I found was anticlimactic in the best way: there&#x27;s almost nothing there.<pre><code> - &quot;AI agents&quot; are just functions the model tells you to call
 - &quot;Memory&quot; is literally just a list you append to and send back
 - &quot;RAG&quot; is search, concatenate to prompt, send it off
 - &quot;Multi-agent systems&quot; are just API calls in sequence
</code></pre>
It all clicked after that. Not because the patterns are hard. They&#x27;re not. In fact, they&#x27;re trivial. They&#x27;re just buried under layers of abstraction that make them seem hard.I created 7 modules showing the basics: API calls, conversation state, tool calling, RAG, streaming, prompt chaining. Each one is heavily commented, nothing fancy. Side-by-side examples for Claude and GPT so you can see they&#x27;re fundamentally the same thing.Now when I use frameworks, I actually know if I need them or if I&#x27;m just adding bloat.Repo: <a href="https:&#x2F;&#x2F;github.com&#x2F;jmedia65&#x2F;learn-ai-right" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;jmedia65&#x2F;learn-ai-right</a>