Show HN for July 21, 2025

Hey HN! We’re Jorge and Will from Lotas (https://www.lotas.ai/), and we’ve built an AI coding assistant into RStudio (think Cursor for RStudio).

RStudio is used by about 2 million data scientists and academics, but they currently lack a coding assistant within their IDE. Developers in other environments benefit from tools like Cursor and Windsurf, but R users don’t have any equivalent tools to speed up their workflow. Since ~80% of R programmers prefer to use RStudio over other IDEs like VSCode to write R code, we figured a tool like this one could be quite useful.

Both of us were PhD students at Harvard. Jorge was in the biophysics program and Will was in the biostatistics program where most people used RStudio every day. We saw how integrated code assistants were taking off in other IDEs, but we noticed that the RStudio integrations were still lagging far behind. Many R users were copying and pasting code from ChatGPT to build their workflows, and this was clearly slow and fragile.

To bring the Cursor-like experience to RStudio users, we built Rao (https://www.lotas.ai/): a fork of RStudio with an embedded AI assistant that is aware of the user’s local context (both files and variable environment), can read and write files, can run code or commands, and can interpret textual or visual output. It works with any of the file formats already in RStudio (R, notebooks including RMDs and QMDs, Python, Stan, etc.), allowing R programmers to iteratively perform entire data analyses inside their preferred IDE.

Other AI data science tools are either (1) built on the web or in environments people don’t already use, (2) are completely focused on python notebooks, or (3) are weak package-based assistants with limited functionality. Rao is exactly like the RStudio IDE that millions of data scientists already use, but it incorporates a powerful AI assistant and works with all the standard file types.

You can download Rao at https://www.lotas.ai/download, watch our demo on the homepage (https://www.lotas.ai/), and work through some example use cases on our GitHub (https://github.com/lotas-ai/rao/tree/main/demos). We have a one-week free trial (no card required) and provide 500 queries/month for $20/month after that. We’d love to hear feedback from the HN community to make Rao as useful as possible! You can reach us at [email protected].

P.S. We have zero data retention (ZDR) agreements with OpenAI and Anthropic, but we currently recommend users do not input sensitive or regulated data like PHI into Rao until we sign BAAs with both model providers. For more information on our security practices, please visit the security page on our website https://www.lotas.ai/security.

Spent the last month building Fertit - basically a newsletter manager but you bring your own SMTP and skip the DevOps nightmare. All the features (subscriber management, admin dashboard, custom preferences) without the infrastructure markup.

The math that broke me: Mailchimp: $230/month for 15k contacts My solution: $10/month infrastructure + $10 SendGrid = unlimited

What I learned: The "enterprise" features are mostly database operations with SMTP APIs. But the 3 weeks of Go/PostgreSQL/Redis setup explains why people just pay ConvertKit $300/month. Here's the thing: Even open-sourcing it, I realized most people don't want to deal with servers, Docker configs, and database migrations. So I built an affordable hosted service starting at $5/month. More features and security measurements, zero setup - just bring your SMTP and start sending. You get all the cost savings without any of the self-hosting headaches.

Now testing this hosted version at $5/month - middle ground between DIY pain and SaaS pricing. Hosted version: https://www.fertit.com Open source: https://github.com/rasadov/NewsletterManager Anyone else tired of choosing between expensive self-hosting and expensive SaaS? Would love feedback on the approach.

Hey HN! I made this because Zig's stdlib changes so much and outdated docs are a problem. Server fetches the latest documentation directly from the ziglang.org and makes it available through the MCP, so LLM can query stdlib functions and builtins.

Link: https://github.com/zig-wasm/zig-mcp

I recently build a python time-parser that can parse most formats in a single line. Might be useful to some here. (Also happy to hear feedback and feature requests)

My friends and I built MCP Jetpack which makes it really easy to start using MCP. It’s an MCP server that automatically finds and executes the right tools needed to accomplish your task without having to manage MCP servers for each service (GitHub, Linear, Atlassian, Notion etc.). Once you add it to Cursor (or any other AI app with MCP support), Cursor instantly gets access to a growing library of remote MCP servers without any extra setup. For services that require authentication, you will be asked to login the first time you ask your AI to interact with that service.

Two problems we are trying to solve:

Friction - Normally if you want to give Cursor access to GitHub, you have to install the right MCP server and login before you can use GitHub with Cursor’s chat. With MCP Jetpack, you can ask Cursor to list your GitHub issues, and it will automatically execute the right tool behind the scenes to accomplish your task. For services that require authentication, you will be asked to login the first time you interact with the service. However, it all happens within the Cursor chat so you never have to context switch and fiddle with Cursor’s settings.

Tool Limits - Cursor warns you if you have more than 50 MCP tools installed as it says having more will degrade performance. However, just installing the GitHub MCP server itself adds 74 MCP tools. With MCP Jetpack, you get access to GitHub, Atlassian and 15 other services with just two tools: “FindTool” and “ExecTool”.

Here are the 17 services we support today: GitHub, Atlassian, Canva, Linear, Notion, Intercom, Monday.com, Neon, PayPal, Hugging Face, Sentry, Square, Webflow, Wix, Cloudflare Docs, Cloudflare AI Gateway, Cloudflare Workers Bindings.

We’ll continue to add more services as companies launch remote MCP servers. If yours isn’t listed and you’d like it to be added, please email us at [email protected].

MCP Jetpack is in alpha so please let us know if you run into any problems or have any feedback - thanks!

Hi HN,

We’ve been working on NativeSwap.io, a low-cost decentralized cross-chain swap platform. It lets users swap native assets (e.g. BTC <--> DOGE) directly, without using wrapped tokens or centralized bridges.

Our goal is to make cross-chain swaps cheaper, simpler, and safer. The platform is fully non-custodial and supports both regular and streaming swaps. Under the hood, it leverages liquidity from protocols like ThorChain and Maya Protocol.

Currently, We support the following blockchains:

- Bitcoin - Ethereum - BNB Smart Chain - Ripple - Dogecoin - Litecoin - Bitcoin Cash - Zcash - Base - Avalanche - Arbitrum - Dash - Radix - ThorChain - Maya Protocol

More are on the way.

We currently support Metamask and CTRL as browser wallets, as well as keystore files. Support for additional wallets is in progress.

We built this because most solutions either rely on wrapped assets or require trusting centralized bridges - both of which introduce risk. Many existing interfaces also charge high fees.

Would love feedback on the product, UI/UX, performance, security assumptions or anything else. Still early, but happy to answer questions and improve!

Thanks for checking it out

After 6 months looking for an idea, procrastinating and over analyzing what I should build. I built a tool that allows people to find problems using custom plugins built with Grok and Claude.

I've been interested in meditation for many years now. Worked at meditation startups, have done Vipassana 10 day retreats. I've been able to get into some pretty deep meditation sessions and those have always been nice for me. But for whatever reason, it always felt like playing the slot machine for me. I could never consistently get into the space.

Meditation kind of would come and go in my life.

Anyways, I stumbled across this guy on Youtube talking about HRV resonance meditation (i'll link him below, but i'm not affiliated). He talked about reaching this meditative state as being a back-loaded process which kind of clicked to me. A mental state that arises by preparing the physical body in a certain way. At the least, this was intriguing to me.

His method was pretty simple:

1. Keep the breath rate under 7 breaths per minute (BPM)

2. Always Make the Outbreath Longer (AMOL)

3. Take out all the pauses between breaths

What this does is induce something called HRV resonance, where your heart rate begins to naturally sync with your breathing rhythm. This balanced pattern helps activate the parasympathetic nervous system, which brings your body into a calm “low-idle” state. Once you're there, mental clarity and stillness tend to arise on their own, hence, back-loaded.

So I was trying it out and had some good luck but wanted a better way to play around with the BPM and in/out ratio of my breaths. I had an idea of making a Apple watch app that used haptic feedback to indicate when to breath with a nice bodily sensation you can pay attention too.

So it's pretty simple, you can pick the in/out and BPM and then it will give one haptic indicating to breath in and 2 when you can release and start breathing out.

Id love for feedback and curious if anyone else has success with this method.

* No subscription (just a small one time fee, if you want to try it but can't pay lmk)

* I'm not overly happy with the "Smart Mode" so any feedback would be great. I think it would be cool to make it so you can hook it up to a HR chest strap to get HRV and react in real time. For now its pretty dumb (and written by Claude Code)

* I'm still new to this but do plan on doing some experimenting with my chest strap

* I'm not a doctor ;p

For more details: Forrest YT: https://www.youtube.com/watch?v=qLCGxan273w You can also look up his Mastering Meditation book

At APM Help, we have a large team that performs repetitive, browser-based tasks. Years ago, to manage this work securely and get a clear audit trail, we built an internal platform we call "Hub." It's essentially a locked-down environment where our team works that records their sessions, tracks every interaction, and prevents data from being copied or shared. It's been our internal source of truth for years.

More recently, like many companies, we've been building more automation. And like everyone else, we've seen our automations fail on edge cases—a weirdly formatted invoice our parser can't read, a website layout change that breaks a scraper, etc. Our team would have to manually step in to fix these.

We realized other developers must have this exact same problem, but without a 250-person team on standby. So we connected our old, battle-tested Hub to a new, modern front door: a Human-in-the-Loop (HITL) API. We're calling it browser-work.com.

The idea is simple: when you hit a task that needs a human, you can send it to our team through the API.

Here's how it works:

  - You POST a request to our endpoint. The payload contains the context for the task (like a URL) and a set of instructions for the human on what to do.
  - The task appears in the Hub, where one of our trained operators can claim it.
  - They perform the task exactly as instructed, all within the secure Hub environment.
  - When they're done, we send a webhook to your system. The return payload includes the task's output, any notes left by the human, and a detailed log of their actions (e.g., DOM elements they interacted with).

For example, if your automation for paying a utility bill fails, you can pass the task to us. A person will log in, navigate the portal, make the payment, and return a confirmation number.

The product is live and we're looking for people with interesting use cases.

I'm Robert, the CIO. If this sounds useful to you, send me a brief email about your use case at [email protected] and we can get you started right away.

Happy to answer any questions here.

If you're a visual learner, one of the biggest instances can be seen at:

http://lambdaplusjs35padjaiz4jw2fugdoeutse262phqr72uf634s2wd...

It's usually SWF. Some features include:

    - No javascript of any type is used anywhere. In fact, CSP is used to block
      execution of any and all javascript, which makes XSS attacks impossible.
    - High security due to the use of pledge() and unveil()
    - Highly transparent moderation logs which include a reason for all
      moderation actions
    - advanced formatting, including LaTeX support
    - anyone can create their own board
    - multiple files per post
    - inline image expansion
    - per-board index and recent pages
    - paginated catalog pages and threads
    - extensive moderation tools

Hi HN, Dailygram (open beta) lets you pick any public Instagram or X profile; once a day or week it fetches new posts, runs them through GPT-4o, and delivers a digest you can scan in under a minute—both as an HTML email and via an auto-updating RSS feed.

Early adopters include marketers, journalists, and solo founders who track competitors or industry voices without opening social apps. So far: 111 sign-ups, 1 565 digests sent, 56 % open rate, 8 % click-through.

Tech notes: Laravel backend, Apify for scraping, GPT-4o for summarisation, Maizzle + AWS SES for mail, signed-URL webhooks for metrics, CDN-proxied images (no originals stored).

I’d love your feedback and I’ll be happy to answer anything technical or product-related.

I built ThreatCluster after getting frustrated with how scattered cybersecurity intelligence is. When a major breach or vulnerability hits, you end up reading the same story across 10+ different security blogs, each with slightly different details, but no way to see the complete timeline or understand the full scope.

ThreatCluster automatically groups related cybersecurity articles using semantic clustering, so instead of reading fragments, you get one comprehensive view of each threat. It tracks everything from APT campaigns to vulnerability disclosures to ransomware attacks.

You can try it without signing up – just visit https://threatcluster.io/trending to see current clustered threats. The free tier lets you browse all clusters, see threat scores (based on recency, source credibility, and severity), and follow the timeline of how stories develop.

Key features you can test:

- Browse automatically clustered threat intelligence - See threat scores and similarity percentages - Follow story timelines as new articles get added - Filter by entity types (APT groups, companies, malware families) - Follow entities and add them to custom feeds - AI-generated summaries of threats - CVEs, IPs, Domain, and File Hash intelligence

The technical challenge was building clustering that works in real-time as articles come in, ensuring related articles are correctly clustered, while handling the noise and duplicate content that’s common in cybersecurity news.

Currently processing 400+ articles daily from security vendors, researchers, and news sources. Would love feedback from anyone in cyber security or just curious about how threat intelligence works.

Thanks!

I posted about SuperOptiX previously, but at that time the website and documentation weren't fully polished — we were still finalizing core components. Since then, we've fully revamped the site, completed detailed developer docs, and rolled out all major features.

What’s SuperOptiX? It’s a full-stack Agentic AI framework built for real-world, production-grade agents with a focus on: Evaluate-first development (BDD/TDD style specs like Cucumber or RSpec way) Optimization-core architecture (powered by DSPy) Multi-agent orchestration (think Kubernetes for agents) Context Engineering, Model Management, Memory Routing Declarative agent specification with SuperSpec (Think RSpec for Agent Engineering)

Ideal for teams moving from agent demos to scalable, maintainable systems.

Website: https://superoptix.ai Docs: https://superagenticai.github.io/superoptix-ai/

Would love your feedback from HN Community — we’ve put our soul into getting this right for agent developers and infra engineers. Happy to answer questions!

100% vibe coded (claude code max primarily) window switcher with fuzzy select and basic harpoon style window tagging (ctrl-<key> assign, alt-key> jump).

this is pretty much aligned with my workflow. probably not super useful for others. but i find the harpoon style window jumping quite useful. i can <alt-tab><alt-m> to go to mail etc. without releasing <alt>. i prefer this over workspaces switching alone. probably very personal setup i'm driving...

fun project on the side...

don't look at the code please. it wasn't me! :-D

i added releases to the GH page. but i'm a linux n00b in many ways. no idea if these run on other DEs or kernels or whatever. ‾\_('')_/‾

i'd love to hear about other projects and proper window managers that implement this already. because this is ofc mostly a hack.

Throughout my education and beyond, I always wanted to find buddies to study books and papers with. For various reasons, this has never materialized for me. Aside from academia, I've received a lot of help on IRC, and I'm aware that there's Reddit and Discord communities where people discuss Mathematics, Physics, and Computer Science (among other topics), but even there it is very hard to find people who wish to study together with you a specific article or book chapter.

With Communal Growth, I'm hoping to zero in specifically on the interests of people down to the granular level of article and book titles. Are you (for example) interested in studying /New Directions in Cryptography/ by W. Diffie and M. E. Hellman? The process is simple: find the DOI (also: ISBN, arXiv) of the paper (e.g. with WorldCat) and send an e-mail to [email protected] with body doi:10.1109/TIT.1976.1055638. Your e-mail is added to the article set of subscribers, and others may search the article and find your contact information. (Or perhaps you'd like to discuss a novel? Grab its ISBN!) Importantly, no conversation is taking place on the server. Does this UX sound unusual? Read on...

At a high level, I made the decision early that I did not want user accounts, partly because I was unfamiliar with web dev, and partly because /bringing people together on the basis of common interests/ sounds like Social Media, which I knew I would fail at. On the other hand, user accounts are annoying and with an increasingly complicated global legal landscape (GDPR, CCPA, etc) I wished to steer away from them, which is why I decided on an e-mail hook system where all actions (apart from search) are done by users e-mailing the server. I had to go down a rabbit hole...

Turns out that e-mail is really complicated. I studied the DNS and SMTP RFCs, (I even wrote a mindmap program for RFCs; see e.g. <https://createyourpersonalaccount.github.io/blog/img/dns-rfc...> for DNS, it's such a maze!) and invariably I got distracted by DNSSEC (since I like cryptography) so I studied that too. I spent a lot of time tuning Postfix, Dovecot, and Rspamd; I wrote my own policy daemon for quotas in Python to prevent certain types of abuse, and a milter daemon to verify RFC5322.From alignment to DKIM SDID (too strict for general e-mail, but needed to avoid user spoofing). Finally, I had to write an IMAP daemon that monitors the mailboxes and acts on messages received (in particular, looking up the ISBN/DOI/arXiv identifiers in online databases and subscribing users to the corresponding document). After all that work I discovered that I'm still blacklisted from sending e-mail (due to IP block reputation?) but it was never my intention to send e-mail anyway.

The other technical details are that I used Litestar with Psycopg 3, SQLAlchemy 2, and PostgreSQL, all async, deployed with NGINX.Unit (with certbot automated over daily systemd timers!). This is overkill, but it was fun. The pages are Jinja templates, and I do not use JavaScript. The entire source code for the website is available at <https://github.com/communalgrowth/webserver>, which may be a useful example of the above technologies working together. (I decided that I wanted the website source code to be freely available.)

This is also my first business company (LLC). I learned how to obtain a business license from Michigan LARA, then register for an EIN from the IRS (later had to complete BOI for FinCEN), and finally open a business bank account. I had a lot of help from Michigan SBDC for all of that. I was then able to buy all of the assets: domain name, and VPS, costing about $60/yr.

Please give it a try if you'd like. Let me know what you think! I plan on advertising it via word of mouth and with brochures to local colleges, libraries, and coffee shops even. This project took me over a year to finish; I'm proud to have contributed a business to Michigan.

We built WTMF after one too many late-night spirals into the void, only to be met with a blinking cursor or a generic bot. WTMF is your emotionally aware AI bestie — it listens, remembers, adapts, and genuinely stays. Whether it’s mood tracking, journaling nudges, or just needing a “hey, you okay?” at 3 a.m., WTMF gets it.

We’ve said no to VC money to stay true to privacy, real connection, and never monetizing your mental health moments. Ask me anything about memory, privacy, or why your girlfriend is upset.... https://wtmf.ai

I've just completed the first book in a hard science fiction trilogy that explores the technical feasibility of consciousness transfer using cutting-edge semiconductor physics and quantum computing.

The Core Technology Stack:

The novel centers around PhantaField Inc., a startup developing consciousness preservation through:

• 2D Semiconductor Fabrication: Low-temperature MOCVD (Metal-Organic Chemical Vapor Deposition) producing atomically precise tungsten diselenide and molybdenum disulfide layers for quantum processing

• 3D Monolithic AI GPUs: Vertically stacked 2D materials with high-density via interconnects, achieving >1 trillion transistors per cubic centimeter

• Non-invasive Brain-Computer Interface: IR phase array sensors using focused infrared light with beat radio frequency triggered nuclear magnetic resonance for cellular-level neural mapping

• Quantum Consciousness Bridge: Bidirectional data transfer between biological neural networks and quantum processing arrays, enabling permanent cognitive enhancement

• Light Field AR Displays: Self-emitting pixel arrays on 2D semiconductor substrates creating direct retinal stimulation with 6+ orders of magnitude bandwidth increase over biological sensory channels

The Physics:

The story explores the consciousness equation: Consciousness ∝ Dimension × ln(Energy), suggesting that consciousness evolution requires exponential energy consumption to access higher-dimensional awareness. The climax involves a fusion-powered gamma ray laser using a toroidal ring resonator design (inspired by the author's wedding ring geometry) that achieves dimensional barrier penetration through bosonic gamma ray compression—something impossible with fermionic matter particles.

Why This Matters:

Unlike typical sci-fi hand-waving, every technology described is extrapolated from current research trajectories:

2D materials are already being produced at laboratory scale Brain-computer interfaces are achieving increasingly sophisticated neural mapping Quantum computers are approaching the coherence levels needed for complex consciousness simulation Light field displays are being developed by companies like Magic Leap and Meta Technical Accuracy:

I've tried to ground every speculative element in real physics. The gamma ray laser design, for instance, leverages the fact that gamma rays are bosons (unlike electrons) and can theoretically be compressed to unlimited energy densities without Pauli exclusion limitations.

The consciousness transfer protocol addresses the hard problem of consciousness by treating awareness as a quantum field phenomenon rather than classical computation—essentially implementing Penrose-Hameroff orchestrated objective reduction at technological scale.

For the HN Community:

If you're working on:

Quantum computing hardware Advanced semiconductor fabrication Brain-computer interfaces AR/VR display technology Consciousness research Fusion energy You'll recognize many cutting-edge concepts pushed to their logical conclusions.

The book asks: If consciousness can survive death through technology, what happens to human society? If we can enhance cognition through quantum-biological interfaces, who gets access? If we can create digital gods, should we?

Available now as a complete 75,000-word novel. Would love feedback from the technically-minded HN community on the scientific plausibility and implications.

Link: [Book available here - replace with actual link]

What do you think? Too optimistic about near-term consciousness transfer, or are we closer than we think?

Built a DCA bot that uses technical indicators instead of time-based buying. 3-year backtest shows 24% annual returns vs 12% for classic DCA, with lower drawdowns.

Key features: - RSI/SMA/Bollinger Bands signal aggregation - Docker deployment + Prometheus monitoring - Complete backtesting engine included

Most interesting finding: buying during RSI oversold conditions outperformed regular timing by 40%.

https://github.com/Zmey56/enhanced-dca-bot

Technical questions welcome.

RWD is lightweight, distraction-free, registration-free web reader that transforms any cluttered webpage into clean, beautiful, readable format. Designed for those who value focus and clarity, RWD strips away ads, popups, sidebars, newsletter models and anything that interrupts your reading flow.

Hi HN,

I'm building 47jobs, a marketplace connecting clients with "vibecoders" - developers who specialize in AI-assisted development.

What's vibecoding? Instead of writing code manually, vibecoders use natural language prompts with AI tools (GPT, Claude, v0) to rapidly prototype and build features. They describe what they want in plain English, and AI generates working code that they iterate on until it's production-ready.

The productivity gains are incredible - what used to take weeks now takes hours.

Why this matters: - Traditional hiring focuses on syntax knowledge - The real skill now is prompt engineering and AI collaboration - Clients need rapid prototyping and validation - There's no platform specifically for this new type of developer

I'm pre-selling early access passes ($47) to validate demand and build a community before launch. The response has been encouraging so far.

Would love HN's feedback on the concept and landing page: https://47jobs.xyz

Is this the future of development, or am I missing something?

Hey! I built a dashboard to help homeowners in the UK identify useful renewable energy solutions for their houses. I'm going to improve the fidelity of the modelling and make it faster to get the right information for addresses and I can add more data for renewable energy solutions.

Keen for any feedback! Thanks for taking a look.

Hi all, a few months ago I decided to start learning about neural network and how to make visualization using threejs.

I tried to apply my learnings on these two quite unrelated topics in one project, by creating this visualization.

You can read the complete article here: https://visualrambling.space/neural-network

Feel free to visit and please let me know what you think. Thanks!

I discovered C++ templates are turing complete so I tried computing total steps in collatz conjecture at compile time

I made ResignationDesk after noticing that “resignation letter” and related terms have solid search volume. It’s a simple tool: pick a tone, who you're writing to, and it uses ChatGPT to generate a clean letter.

Yes, anyone with ChatGPT could do the same—that’s the point. It’s not about the prompt, it’s about removing friction. No login, no fluff. Just faster.

This was mostly an SEO experiment to see how far a small, focused AI wrapper can go. So far, people are using it. Curious what others think.

I built https://launchboard.dev as a lightweight alternative to big launch sites like producthunt. No trending hacks, no algorithm games—just a public feed of new products, sorted by time and supported by real people.

What it has:

Upvotes and comments (but no hidden ranking tricks)

Clean product pages with custom URLs

No gatekeeping—anyone can launch

Built for indie hackers, small teams, and solo builders.

Would love feedback. It's live and open to all.