Hacker Newsnew | past | comments | ask | show | jobs | submit | Leftium's commentslogin

Lisp is often used to create domain-specific languages.

- Lisp S-expressions are powerful because they represent both code and data. So data is code; code is data: it is very simple to augment the language.

- Lisp S-expression syntax conveniently mirrors the tree shape of Abstract Syntax Trees.

F# also supports macros and meta-programming: futures (F# version of async/await) were added to existing versions of F#. No need for a new language runtime!

If JS had that level of meta-programming, I wouldn't have to wait for pattern-matching support! And pipes!


I use a built-in MacOS wallpaper called Solar Gradients[1]

It looks very similar!

[1]: https://youtu.be/0mf8YaWN5qE?t=1m21s


Scott Tolinski is sort of an expert with sync engines, and he just compared several solutions here: https://youtu.be/1uVR5X7HpI8


Best Realtime Sync Engine (2024): https://robotist.com/realtime-sync-engine

Theo Browne (T3) detailed the different methods he used to sync user data for https://t3.chat: https://youtu.be/3gVBjTMS8FE

He looked into several solutions like https://zero.rocicorp.dev, but decided to roll his own.

He finally decided to let Convex handle sync for him:

- https://youtu.be/xFVh9beupwo

- https://youtu.be/gZ4Tdwz1L7k


This successful B2B SaaS was (originally) developed by a single person who didn't know how to code (without any investors): https://www.paperlesspipeline.com

The key insight was: marketing was the most important part. And the most important part of marketing was finding (and validating) a painful problem businesses would pay for.

- It is important to talk directly with real prospects: real business owners. You can use stuff like Gartner reports to help find a niche, but it is essential to talk directly with real people who have the problem.

- Validation involved getting a prepayment even though the software did not exist. Not even a design. Prepayment got the initial 10 or so "champions" a lifetime discount and personal input into how the app would be developed.

- The prepayments served as a much stronger validation than a verbal "yes." Also it funded development of the initial version. The 10 champions' input also made the initial version much better.

- Generally, a "painful" problem will result in at least a 10X ROI (additional revenue, saved expenses, saved time, etc). If you describe the painful problem better than the prospects themselves can, they will assume you have the solution and jump straight to "how do I pay you? (sign up)"

- The market research math is roughly: find a niche with at least 10,000 prospects. Plan to capture 1% of this market, or 100 customers. If you can find a painful problem whose solution is worth paying $100-200/month, you should be able to achieve $80K profit/year.

(The Paperless Pipeline origin story is slightly different. For example, there was really only a single true "champion." But this champion was the happy customer of another SaaS that was built with the same process. This champion kept asking for this other SaaS that turned into PaperlessPipeline.)

---

So I think the resources below provide the most risk-free (highest chance of success) way of achieving your goal (I think YC targets "unicorns" with much higher potential ROI with the understanding at least 90% will fail):

- The founder of Paperless Pipeline has been teaching his method: https://startfromzero.com (Specifically "Idea Extraction" is the most important part)

- Apparently he is going to start a YouTube channel where he teaches his stuff for free.

- https://30x500.com is another great course that teaches a similar method, and more geared toward engineer-types.

- https://shop.stackingthebricks.com: Kind of like the a-la-carte version of the 30x500. The two most valuable parts are probably "Sales Safari" and "Just F*ckin Ship."


Veneer: a thin facade over Google Forms/Sheets (https://github.com/Leftium/veneer)

Some samples:

- https://veneer.leftium.com/v/s.1o5t26He2DzTweYeleXOGiDjlU4Jk...

- https://veneer.leftium.com/v/s.1pk4C9jFI02CnZaxo9obsD4oAmLla...

Vercel is ending support for Node v18. Instead of updating my old app, I decided to finish the rewrite of the better version. The old version currently powers this site: https://viviblues.com

Compare to the new version:

- https://viviblues.com/pretty/sheet?u=https://docs.google.com...

- https://viviblues.com/pretty/sheet?u=https://docs.google.com...


My app also shows yesterday's weather: https://weather-sense.leftium.com

- It uses open-meteo's API. Before open-meteo, I tried with OpenWeatherMap's API.

- However, besides requiring an additional call for each previous day's data, OpenWeatherMap's One Call API only supported local time zones for current and forecast data. So depending on the timezone and time, there could be a big gap in the data that needed to be filled with timezone math (and extra API calls. The original DarkSky One Call API did not have this issue.)


I wanted the same thing, so I built: https://weather-sense.leftium.com

- I even blogged about this over 10 years ago: https://blog.leftium.com/2013/12/how-to-display-temperature-...

- Previous version used to show two historical days: https://github.com/Leftium/ultra-weather

- The current version is capable of showing the previous 90 days. I plan to add a weekly overview with two previous days.

BTW the https://open-meteo.com API supports getting historical and forecast data in a single call. All other API's require at least two calls.


I'm working on a weather web app that has world-wide radar coverage: https://weather-sense.leftium.com

The radar data is from the RainViewer API: https://www.rainviewer.com/api.html

- Hmm... it seems this API is also being limited/discontinued...

- I think past radar will continue to work, but only to zoom level 10/7.

- And the future 30-90 minutes of predicted radar will no longer work.

My project was inspired by: https://merrysky.net (And the OG Dark Sky UI)


Two similar Show HN projects:

- This python one is more amenable to modding into your own custom tool: https://hw.leftium.com/#/item/44353447

- Another bash script: https://hw.leftium.com/#/item/41473379

---

They all seem to be built on top of:

- yt-dlp to download video

- whisper for transcription

- ffmpeg for audio/video extraction/processing


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: