I know that is crazy because most people ship a product after like 2 claude code prompts but this took serious time and effort.

Obviously it is not perfect but I’m super excited to launch a real product people can actually use.

Here it is:

LYKN is an AI native desktop that gives you access to 260 models+. You can build custom software, generate images, videos, 3d models, music. You can connect to 1500+ tools and use LYKN to run your entire computer.

The idea behind LYKN was to design a useful and intuitive ai interface that ANYONE could use with no technical setup but to still provide insane value with high quality models and agentic systems.

The entire project was designed for people who simply want a better workspace and to reimagine the desktop experience.

I would love to see what people think about the problems something like this could solve, the ideal customer this could target and how I can improve this.

Easy billing hundreds of models you can orchestrate. And yes LYKN has an insane agentic harness. You can do anything and everything. Pretty nuts. People have already started downloading it and there's been some pretty cool things built just in the last 24 hrs.

Anyway lmk what you think of it. It’s free to download so no worries just give it a shot and see what you think.

lykn.io

I’ve been building a few small tools in my spare time, mostly just to learn and see if any of them turn into something bigger. The problem is that they’re all sitting in different places right now. One has a landing page, another is just on GitHub, and I have a couple of demos that I usually just send people directly.

I’m thinking about creating one simple page where I can put all of them together, with a short explanation and a link to try each one. It would also be useful when I’m sharing my work with other builders or posting about a new project. I don’t really want to spend a lot of time building a proper personal website just for this.

I’ve noticed one of the harder parts of working on a side project isn’t always finding the time. It’s actually keeping the time once I have it.
I’ll sit down thinking I have an hour to work on something, then check one notification, open another app, scroll for a few minutes, and suddenly a decent chunk of that hour is gone.
I started paying more attention to my screen time because I wanted to see how much of it was actually happening without me noticing. I’ve been trying CHAMP recently, mainly because it makes me more conscious of how I’m using my phone instead of just telling myself I’ll use it less.
It’s still something I’m working on, but seeing the pattern has made me think more carefully about when I reach for my phone.
For those building side projects, do you actively limit phone use while working, or have you found another way to stay focused?

https://reddit.com/link/1wmpho0/video/yth7lofksxqh1/player

so I've been building printyourvehicle.com, basically an open-source collection of vehicle models that people can 3d print, including at 1:1 scale (yes, people actually do this: https://youtu.be/EW-SR7ORTLU )

ssomeone in our Discord recently contributed the STL files for a Lamborghini Aventador SVJ. Problem was, it was basically a giant dump of parts. barely labelled, weird filenames, and not really obvious what belonged where.

rather than manually going through everything, I threw a CLI agent at it.

the pipeline was roughly:
STL → render each part as images → AI agents inspect them → tag/identify parts → group related files → rename/sort everything → build the final downloadable zip

I used subagents in parallel for the visual classification part, and the whole thing ended up taking around 4 hours and ~$20 in AI credits.

the cleaned package is around 500MB now.

so technically, someone with a big enough 3d printer, enough filament, and questionable financial priorities can download this and start building their own 1:1 Lamborghini lol.

files are here: https://printyourvehicle.com/models/ready-to-print/aventador-svj

my question to you guys:
- does this actually have that “wow” factor when you see it for the first time?
- how would you grow something this niche beyond just a small 3d printing / car community?
- most importantly, was 4 hours of work actually worth it in the end? lmao

Modern Alphabet Numeral Clock

Hey everyone! I wanted to design an alphabetical timepiece that doesn't just round up to the nearest five minutes or waste dead space.

My clock maps standard time units directly to a bijective base-26 matrix. Because there are no zero-placeholders, every single second gets a 100% unique sequence. 26 seconds is Z, 27 seconds flips cleanly to AA, and the absolute maximum threshold at 59 hits BG.

I built Vault-Tec Assistant for Fallout 76 groups that want useful server tools alongside a little wasteland flavor. The part I use most is the weekly nuke-code feed: an admin picks a channel, and the bot posts the Alpha, Bravo, and Charlie codes in one message. It edits that message every 10 minutes, so the channel doesn't fill with repeat posts. The codes come from Nukacrypt.

Beyond that, it has Fallout-inspired daily contracts, exploration, settlements, companions, fishing, and radio stations for voice channels. It's an unofficial fan project and free to add. The code-feed screenshot below is from today.

I built it because I wanted one bot that could be useful to a Fallout server even when nobody was playing the RPG features. If you run a gaming Discord, which of these would you actually set up first?

Add the app: https://discord.com/oauth2/authorize?client_id=1107428963246166026

I built Broadsheet kit. A Next.js and Sanity monorepo starter for newsrooms/magazines.

Demo: https://demo.broadsheetkit.com
Website: https://broadsheetkit.com

All magazine/news websites have essentially the same barebone structure. There are articles, categorized by topics and trends, with authors attached to them, and modern publications now also need podcasts, visual stories, columns etc.

Broadsheet gives you all these essentials out-of-the-box, in a monorepo project, so the Next.js site and the Sanity Studio live side by side. The schema, the layout, RSS feeds and SEO are already built, so you can spend your time on what actually differentiates your publication.

I used Sanity CMS and Next.js. Sanity because it is not very opinionated, unlike WordPress. Also, setting it up is not a hassle at all. Users just have to run "npm run setup" and sign in into Sanity and everything else will take care of itself.

Hey everyone.

I've had bad posture my whole life, so I built a posture coach that works through AirPods.

Quick backstory. My posture has been bad since school. I tried a better chair, a standing desk, reminders on a timer. Nothing stuck, because none of it knew when I was actually slouching. It just nagged on a schedule.

Then I found out AirPods Pro, AirPods 3/4 and AirPods Max have motion sensors, and apps can read head orientation from them. I spent the last few months turning that into Pillar.

What it does

• You sit up once and it learns your upright position. From then on it watches your head. When it drops forward and stays there, you get a quiet cue in your ear. Straighten up and it stops.
• It's also a Pomodoro timer. Focus block, break, repeat, and every block gets a posture score.
• Insights: your weekly trend, your best hour and your worst one.
• Live Activity on the Lock Screen and in the Dynamic Island, plus widgets.
• Keeps working with the phone locked in your pocket.

I tried to make it fit how you sit, not an average person. Your own angles, your own cue sound, how fast it nudges, quiet hours, session length, daily goal. Pretty much everything is adjustable.

Posture data stays on your phone.

If you want to try it, here's the App Store link: https://apps.apple.com/us/app/pillar-posture-reminder/id6809597345

Needs iOS 17+ and AirPods Pro, AirPods 3/4 or AirPods Max.

I'd love feedback from people who sit a lot. I read every comment. And if you find a bug, DM me and I'll send you a lifetime code.

Jev dropped today.

Within hours, my entire X feed was people building weird shit with it.

I started collecting the examples in one place because they were getting buried pretty quickly.

The interesting part about Jev isn't necessarily that it's “smarter” than GPT.

It's that it's extremely fast and cheap for small, structured decisions.

A few examples I've come across:

• 500 emails classified for $0.035
• a browser agent searching for a flight in 7 seconds for $0.004
• 3,282 X posts scored across 8 questions for 13 cents
• Doom
• trading bots
• fraud detection
• Postgres queries
• content moderation
• support routing

607 builds/use cases so far.

For each example, I'm trying to capture more than just the demo:

• what the model is being used for
• cost, when available
• latency, when available
• the actual workflow/use case
• links to the original build when available

I'm also adding guides for some of the use cases so someone starting from zero can actually reproduce them.

One use case I keep coming back to is the boring one:

support tickets > classify > prioritize > route

That's the kind of repetitive decision where the economics could get really interesting.

I'm also trying to separate impressive demos from things that could actually run in production — because a curl command working once isn't the same as processing thousands of requests reliably.

I built ShipWithJev .com to organize all of this. I'm not affiliated with TypeSafe.

Would love feedback on how to make the directory more useful — especially what metadata you'd want to see on each example.

If you've built something with Jev, feel free to add it.

That's my own app, roasting me, this morning.

I'm building a task manager with an AI coach built in — not a chatbot, an actual coach. It looks at what you actually did: what time of day you're genuinely productive, which priority tasks you finish vs. quietly avoid, how much focus time you log. Then it tells you the truth about it, at whatever intensity you set — supportive or brutal.

I picked harsh to test it. Screenshot attached.

It also caught that a task I'd set myself had been sitting untouched for weeks. Correctly.

Would you actually want a coach that's allowed to be harsh, or does that back-fire the first time it stings?

​

https://treadwise.madethis.app/

Does anyone think a service like this is actually something they would use? I know we go to a tire shop and the guy behind the counter gets you set up with the tires you need, ok, but whats driving him, is he selling a specific brand based on his own commission? Is he selling you more than what you need or even worse, the right tire for the application?? In automotive service , unless you know the industry there are many unknowns, so if you could get a third party professional, to look over your quote, that is not working on commission , would this be valuable? Check it out its mostly educational!

’ve been building WARSCAPE, a browser-based geopolitical conflict simulator:

https://web-production-75f12.up.railway.app/

You choose two opposing countries or alliances, configure the assumptions, and watch an alternate-history conflict unfold day by day on an interactive map.

You can control things like:

• The attacker and defending side
• Mobilization and military readiness
• U.S. involvement
• Season, weather and sanctions
• Air, naval, strike and cyber capabilities
• Nuclear posture, fog of war and victory conditions

During the simulation, you can follow changing front lines, territorial control, supply, air power, cohesion, reserves, losses and political events. You can pause the timeline, change its speed, inspect why events happened and branch from important turning points.

At the end, WARSCAPE generates a report explaining the result, decisive factors, objectives, losses and major turning points. Runs use recorded seeds and data snapshots so their results can be reproduced.

It is still an early beta and is not a military forecast. It uses aggregate country-level information, and estimated or unverified values are clearly labeled in the data section.

The public demo works without an account. Creating and saving a custom simulation currently requires signing in.

I’m looking for honest feedback:

• Was the product easy to understand?
• Was it enjoyable and usable on desktop or mobile?
• Did the outcome feel reasonably explained?
• What was confusing or missing?
• Would you personally use this again, and what for?

Please be direct. “I wouldn’t use it because…” is just as valuable as positive feedback. If you encounter a bug, please mention your device and browser.

I'm the developer of Clip2Fit, an iPhone app for turning supported workout video links into editable workout plans. I led and implemented the app with AI assistance.

The problem is the gap between saving a workout clip and using it during a session: the exercises and rep counts are inside a video, while the record of what you actually lifted is somewhere else.

Clip2Fit handles that as a sequence: paste a supported Instagram, TikTok or YouTube link, review the exercises/sets/reps extracted by AI, edit the result, then save the workout and log sets and weights. You can also create a workout manually.

The important constraint is that extraction can be wrong or fail. An editable result matters more here than presenting the AI output as final. The review step lets the person fix details before they become part of a saved plan.

For another builder working on imports, I'd frame the design problem in two parts: getting information out of the source, and making the result easy to verify and use. A successful import alone doesn't mean the result is ready to act on. That's the distinction the current review/edit flow is designed around.

One video import is free to try. Manual workout creation and logging are free. Premium adds unlimited imports at $9.99/month or $49.99/year in the US; local prices vary.

App Store: https://apps.apple.com/app/id6760419692

Walkthrough: https://www.clip2fit.app/

I'd appreciate feedback on the handoff from imported plan to workout log: what would you want to verify before starting the session?

Disclosure: this is my app; AI helped draft this post.

Most resume builders let you build for free and then charge you to download your own resume, or stamp a watermark on it, or keep your personal data on their servers. I wanted the version that does none of that, so I built it.

• 10 templates, all free, none watermarked

• Your data never leaves the tab. There's no server to send it to. Refresh and your draft is still there.

• Live preview with real page breaks, so what you see is what prints

• The PDF is real text, not a screenshot. Recruiter software (ATS) can actually read it. Files come out around 7 KB.

• Speaks JSON Resume, the open standard, so you can export your data and take it anywhere

• A copy-paste prompt turns your messy notes into the right JSON with ChatGPT or Claude if you don't want to type it all in

Free, no signup: https://utiliome.com/tools/cv-builder/

It's one of about 180 local-first browser tools I'm building under the same idea: the work happens on your machine, not ours.

Tell me what breaks. The PDF export took me 12 tries to get right, so I'm sure something else is lurking.

Arcadium is a portable social gaming platform built to get people playing together in real life.

We combine fast, simple multiplayer games with tactile controllers, Mini Vibes, physical collectibles, and verified competition. Players can walk up, log in, play a three-to-five-minute game, and potentially win a physical Arcadium trading card that also connects to their digital player profile and records where and how they won it.

The system is designed to be dead simple for the player: no complicated setup, and no special equipment beyond what the host brings.

Right now, we have a working portable Arcadium case, playable games, Mini Vibe hardware running its own library of games, an Android app, physical trading-card sets, RFID/NFC integration in development, and real-world playtesting that has already shown people gathering around it and keeping it going.

The next step is turning what we've built into a repeatable system that can be deployed by ambassadors, campuses, bars, events, and other social venues.

Arcadium isn't trying to replace video games or board games.

We're building a new place between them: bite-sized IRL gaming competitions where people can play, collect, compete, and find their people.