HNHacker News
TopNewBestAskShowJobs

hantusk

242 karma · joined March 12, 2013

submissionscomments
hantusk··on Replit's new AI Model now available on Hugging Face
check out Daniel Gross's https://github.com/danielgross/localpilot
hantusk··on Where did the long tail go? (2022)
The Pareto principle is about the converse as well
hantusk··on Nocodb: Turns Any MySQL, Postgres, SQLite into a Spreadsheet with REST APIs
https://github.com/jeffknupp/sandman2 works with many different backends including sqlite, but postgrest definitely have the auth flow and security model right for these things.
hantusk··on Get in zoomer, we're saving React
When reading through the comments here, I really feel like the article was misunderstood. My summary of the article is:

Point 1: React solved all the right things, but its current trajectory, does not prioritize developing the fundamental tooling we need. React does not allow us to build a new Figma (consistent undo/redo in collaborative settings, immediate low-latency mutation of app state to reflect user changes and building fundamentals to SaaS interoperability)

Point 2: People are overwhelmed by the work, and too timid to target the fundamental architecture we need to solve these things. We're just going with the flow.

He is leading https://usegpu.live/ to build the fundamentals right, an encourage people to either:

- help out there, if it fits your use case, or

- start working on targeting these fundamentals, where most needed for your own use cases, don't just go with the flow.

hantusk··on Ask HN: Does anyone else think SQL needs help?
Postgres can do the"function over (partition by col_a order by col_b) as well.

Agree, SQL can already do what the OP proposes in almost the exact same syntax.

and together with CTEs, intermediate temp tables or views, you can totally avoid the spaghetti and break down the queries in manageable chunks.

hantusk··on The future of Nginx: Getting back to our open source roots
Hm, i found this to be pretty straight forward through Docker:

e.g:

    FROM caddy:2-builder AS builder
    RUN xcaddy build \
        --with github.com/greenpau/caddy-security

    FROM caddy:2
    COPY --from=builder /usr/bin/caddy /usr/bin/caddy
You can keep your container image, and just do the "COPY --from=builder" step in there. We do this in CI/CD on every build.
hantusk··on Scorched Earth: The Mother of All Games
Me too!

It had a cheat code, so if you set the power to a certain amount it would shoot through the grass

hantusk··on I'm all-in on server-side SQLite
Same pattern is ActorDB: https://github.com/biokoda/actordb
hantusk··on Malloy – A Better SQL, from Looker
Not a big fan of SQL dsl's personally. If you feel you'd need a better query syntax and plotting for interactive use, I'd suggest looking at the dataframe libraries that have a larger community to discover how to solve things and have ironed out issues.

Specifically dataframe logic is easily mappable to dbplyr https://dbplyr.tidyverse.org/articles/sql-translation.html for those that know R and ibis https://ibis-project.org/ for Python/Pandas

hantusk··on HashiCorp IPO today
Gitlab just IPO'ed this year. It seems unlikely, that this would be a motivation.
hantusk··on Dask – A flexible library for parallel computing in Python
a path I haven't explored yet is Koalas, which touted a 80% API compatibility with pandas at version 1.0. Now its at version 1.8, and it has the pandas API as the main design principle for function interface design.

https://databricks.com/blog/2020/06/24/introducing-koalas-1-...

hantusk··on Comparing SQLite, DuckDB and Arrow with UN trade data
this benchmark is more comprehensive for this type of analytical work:

https://h2oai.github.io/db-benchmark/

hantusk··on Adventuron – Text Adventure Authoring in the Browser
we're not far away :) https://twitter.com/nshepperd1/status/1446004530433323018

There's a notebook too, if you're curious

hantusk··on Python Best Practices for a New Project in 2021
Does it not work with miniforge/mambaforge? (https://github.com/conda-forge/miniforge)

I use this with my m1 mac, and it works great.

hantusk··on Excalidraw+
I hope kinks have been worked out. z-index errors, sync not working for some participants and JavaScript alert windows about losing data, were issues the three times I've used it with my team, so i basically dont dare suggest my team using it again.

They've been very good at communicating in the GitHub issues I've filed, so hopefully it works for others, but I would advise you to test with a few co-workers before facilitating anything in a group

hantusk··on Practical SQL for Data Analysis
Agreed. https://ibis-project.org/ and https://dbplyr.tidyverse.org/ can compile dataframe-like input to SQL, which might bridge the gap in tooling (although there still are small differences to the pure dataframe syntax)
hantusk··on Ask HN: What diagrams do you use in software development?
I use mermaid: https://mermaid-js.github.io/mermaid/#/

And excalidraw: https://excalidraw.com/

hantusk··on MLJ.jl: A Julia package for composable machine learning
I would go for decrappify: https://www.fast.ai/2019/05/03/decrappify/

GAN's are hard to train for a new ML practitioner, VAE can have blurry results if you are not using skip-connections like in U-NET.

It sounds like you have training examples (images pre and post processing), but would benefit from an approach using transfer learning since you don't have that many.

Good luck!

hantusk··on SQLAlchemy 1.4
Neat! you could prototype out the db in sqlite, and then have sqlalchemy introspect and write out a db-agnostic model from this.

Happy to see the support around foreign key relationships and lookup tables here.

Thanks!

hantusk··on SQLAlchemy 1.4
Yeah, see: https://dataset.readthedocs.io/en/latest/

It doesn't do foreign relationships though, guiding you to use sqlalchemy when you reach that level of complexity. It works for all databases supported by sqlalchemy though

hantusk··on SpaCy 3.0
This. Wholly agree. Currently running a large labelling task with 12 labelers.

Using Amazon ground truth which works fine (although seems quite MVP outside the core functionality e.g wrt reporting or user creation).

What tool have you had success with?

hantusk··on PixelNeRF Neural Radiance Fields from One or Few Images
.. and your phone already does this: https://www.google.com/amp/s/www.macworld.com/article/344463...

Also, im sure I saw a very nice video on this, with a car driving through the landscape, using previous frames as well. Can't find the link though :-\

hantusk··on Airbnb S-1
Wow, totally opposite my view. As a family we always want to stay somewhere with a kitchen and a private lounge area - hotel suites are cost prohibitive to this, and a lot less charming.

When finding chalets or cabins/lodges for outdoor trips, Airbnbs are often much better located.

To me it is not so much hotels they need to replace, but a streamlining of booking vacation houses to the point where it is almost as convenient as hotel bookings are, but with vacation house amenities.

hantusk··on Old, Good Database Design
Just use TEXT in other databases as well. It really shouldn't matter much in modern dbs
hantusk··on Cozy is a modern audiobook player for Linux
Audiobooks in Kindle is at linked to the text, so its easier to skim pages for where you dozed of. Best solution I've found so far.
hantusk··on The battle to invent the automatic rice cooker
Also, don't add salt until the process is finished, as that also makes it more starchy
hantusk··on Silicon MacBook could launch on 27 October and cost $800
Isn't it possible to buy most high-end smartphones with 256 gig of RAM?
hantusk··on Dexplot: Python library for data visualization
I'm really fond of https://altair-viz.github.io/ which wraps and translates into https://vega.github.io/vega-lite/.

The interactivity and cross-filtering possibilities is really good when you have many datapoints.

From vega-lite I miss that they did not develop object constancy (https://bost.ocks.org/mike/constancy/) animation possibilities since they're building on top of d3.js.

hantusk··on Show HN: Summarize your time availability in seconds
like https://doodle.com/
hantusk··on Vercel, formerly Zeit, raises $21M Series A
^ The re-brand makes much more sense in this light - a young company that can still only focus on 1 product, needs a name that is associated with their current product, and not now v1 and v2.

But that was not the reason given - strange to cause such confusion.

← PreviousPage 2 of 4Next →