earonesty

agreed. ai-generated seo will kill your site too!

is this the similar to https://arcanon-ai.vercel.app/ ??

Mostly I did this because I have a lot of dynamically generated pages, where the actual structure of the page is somewhat controlled by the user. I also have a lot of small sites that I don't want to pay an enormous amount of attention to that need good OG images so that when people post URLs they don't suck.

I will say that for some sites the images will be only ok (stripe was only on). But as long as you have a good mobile render, It will work very well.

If you don't know what OG images are it's basically the miniature version of your site that shows up on social networking sites like Facebook and X.

Some people spend a lot of time optimizing that sort of thing but you can just throw this in your meta tags and it works well enough for now.

Anyway I would welcome some feedback, it was sort of fun to put togethe.

Small calculator to model equity splits based on measurable contributions like:

• hours worked • cash invested • revenue generated • expenses covered

Calculates “ownership slices” from those inputs and shows how the split changes when assumptions change.

Nothing you can't do in excel. The full site (also free), allows more complex calcs.

This only applies to advertising. Not long, manipulative conversations with salespeople. Which is what happened. Any questions about this were conveniently deleted from the subreddit too.

This is faked. There are no details in the post about how the bot was set up. Clearly there are no e-commerce systems that have MCP tooling that allow arbitrary 80pct discounts without explicit permission from the owner.

Any advice or requests for more details are deleted by the moderator.

This is a "set up", designed by the owners of the subreddit to make people who don't know how AI and tooling work afraid of AI. Clearly the author is colluding with the moderators to prevent difficult questions or concerns.

Upload a job description and up to 200 PDFs or DOCX files. CV Match returns:

    Embedding + tsvector scored ranking (OpenAI)
    GPT-4.1-mini reranking + 5-bullet summaries
    CSV export

No signup required to try it — free for up to 200 resumes/month. Tested it on batches of 1,000+, and it holds up (worker/scale/queue).

Stack: FastAPI + async workers, GPT-4.1 rerank, pgsql, Fly.io, nltk

Would love feedback on:

    Ranking quality
    UX flow
    Whether $49/mo feels right or not for pro-mode

I use it myself, because I wanted to control the prompt and the model used. The "service version" lets you... set the "focus" and pick from claude or openai (claude is a little nicer imo).

This is not about posting some big summary, it's about creating a draft for an actual human to pick through.

It drives focus and productivity, doesn't take away from it (for me at least)

Anyway, a lot of it runs in the browser and uses indexdb for storage. Should be possible to make all of it do that.

Basically nobody but me and maybe you and few other people right now. I'll post back here in a week or so and let you know. The graphs on my little logging manager aren't great I might add the Google analytics thing. Although that kind of cookie stuff bothers me.

All the other sites that do this seem to be full of malware, popunders, spam and other annoying stuff.

I might throw ads on it one day, but it took me 2 hours to build it, and it runs clean.

Any advice for features? Was thinking of linking it to a crop/cut thing.

At the office, I find myself frequently wanting to describe system processes, class organization, or business processes in a flowchart. So I made a little site that links an llm to a charting engine for my own personal convenience.

I added a login button so I don't get too much spam and I can throttle use (for now). Looking for advice.

My goal is:

- one input box. maybe a browse button to pick a file - no controls, sliders or things. just free form text as the only ui - currently it's only really good at flowcharts and system diagrams - want suggestions for types of charts people might like, types of data, etc. - vague idea that a "public" checkbox could feed community charts through to some social page where people can comment. maybe just tweet it out. not sure if that idea is useful or annoying to anyone. maybe i should scrap it. - thinking i will do venn diagrams next because it's cute and fun. or maybe maps. or maybe a stock price link? not sure what people want tbh. - generally just want feedback on the idea

I know this could be seen as "yet another object manager", and it is. Thing is, I have a real issue with typical object managers and their interaction with an organization.

An ORM layer should always serve as your primary object interface, not just a weird way to execute SQL calls, while your real objects do the work.

Also, I like my DBA's to be DBA's: to know everything about how to optimize databases, indexes and to do so with SQL. I don't need or want them to be experts at some ORM.

omen2 takes a unique approach to object management by generating "convenience code" based on an existing database's schema. Also, the codegen layer is completely optional.

I'm not going to claim it's done the best possible way, but it's good enough for the dozen or so databases we're using it for, and it's makes far fewer assumptions and has no built-in limitations.

omen2 allows a full separation between marshalling, serialization and databased design.

I find full-blown orm's both restrictive and clumsy for smaller projects. But also i like to use a base layer that allows me to connect to different db's with a unified syntax.

The syntax/structure below seems pretty clean (suggestions welcome), and because it doesn't do that much, it seems unlikely to stop working randomly.

    db = notanorm.SqliteDb(":memory:")
    db.query("create table foo(bar)")
    db.insert("foo", bar=22)
    db.select("foo")
    db.select("foo", bar=[22,33,44])

(that last statement is a placeholder-generated IN clause)

It's left as an exercise for the reader to structure create statements properly (mysql and sqlite syntax is actually pretty compatible).

But realistically, your DBA doesn't want you running create statements from python anyway.

Many libs I write wind up being discarded eventually or replaced with something better. This one has stuck around for 4 years now, and seems to be pretty useful.

It currently only has mysql/python drivers but writing a new driver is pretty simple.

....and someone rolled back your edits. because it's not peer reviewed.

The cloudsync module at pypi can be used from the command line to run a daemon which monitors files for changes and synchronizes them between providers.

For example: `cloudsync file://home/me/gdrive gdrive://folder`

In addition the cloudsync python module can be installed via pip, and the library can be used in software (LGPL) that performs file synchronization. This also provides a unified interface for working with cloud files.

By design the interface is the minimal needed to support sync across a broad set of providers. Metadata that is not needed for sync is not supported.

Support for new cloud providers can be easily added by anyone producing a module that uses the "cloudsync.providers" plugin endpoint. As long as the 'provider tests' pass, any cloud service should work with the sync system.

We're currently supporting the local filesystem (via watchdog), along with OneDrive, GDrive, Box and Dropbox.

This library is probable most similar to rclone, which is much more mature for those who want a single-direction 'rsync' like service without support for events.

If you're over 40, and you cough, there's a chance you microtear your lungs. If you cough up blood for weeks, yes... you will increase pulmonary fibrosis (fancy word for lung scars). That's true for every flu you get... but this one is way harder on the lungs than any flu I've ever had.

I'm having mild symptoms, but it's been weeks, fevers come and go, and honestly it's like nothing I've ever had before.

That's not even true. I've had pain breathing for weeks, they still won't test. They prescibed corticosteroids, antibiotics and bed rest. They won't test until you're nearly dead. the CFR here is wildly overestimated.

One of the biggest reasons the rate may be too high is the method of attribution. For example, in Italy, post-mortem testing of people who were origially presumed dead of other illnesses accounted for a large portion of testing. When it turned out they had COVID-19 at the time of death, the cause was attributed to COVID-19 regardless of comorbidities: https://www.bloomberg.com/news/articles/2020-03-18/99-of-tho...

In other words, many of those who died today would have died within the year. Many were already in and out of hospitals for serious diseases.

So, not only is the IFR wildly unknown without serology tests. The CFR itself is also severely biased.

Currently in the U.S. you cannot get tested unless you have severe, life-threatening illnesses.

The fact that the U.S. has such a low CFR speaks very well of healthcare in the U.S.... considering the majority of those tested have severe symptoms.

Dual licensing is your friend. Look at QT for an example of a company that has managed to generate decent revenue while maintaining a popular open source project.

"Small companies don't have the resources"... that's some backward thinking right there.

By failing to contribute, small companies cost themselves money very quickly. How can a small business afford to live in a world of un-maintainable forks, patches and tech debt?

Only large companies can afford to maintain private forks of FOSS. Small companies have to give back, or they will simply disappear.

Library comparison: https://github.com/cblp/yaml-sucks

Prove it.

can you post a php example with sanitized params? https://github.com/earonesty/liteapi ... easy enough to pull the change.