AskHN Digest

1. Ask HN: What is the most useless project you have worked on?

Top comment by dang

All (since it's been a while): there are about 700 comments in the current thread. To read them all, you need to click More links at the bottom of the page, or like this:

https://news.ycombinator.com/item?id=39942397&p=2

https://news.ycombinator.com/item?id=39942397&p=3

2. Ask HN: What Are You Learning?

Top comment by solardev

Working my way up to calculus for the first time in my life, at nearly 40 years old. I've always hated math :(

When I was a kid, they always told me math would be super useful, especially if I liked computers. Well, 20+ years of a dev career later, I still have never used anything more than basic arithmetic and rudimentary algebra (to calculate responsive component sizes). But with web dev jobs going the way of the horse-drawn wagon, I figured it was time for a career change. Hoping to get into (civil/environmental) engineering instead, but I guess that field actually does use math, lol. We'll see how it goes...

In the meantime, also taking singing classes at the community college, and enjoying THAT way more. We performed at a nursing home a few weeks ago, and that brought SO much joy to the audience there, even though we're just a bunch of amateurs. It's just such a different reception than anything I've ever seen as a dev. Tech rarely inspires such joy.

If I could start all over again, I wish I would've pursued music over computer stuff. Much harder life though!

3. Ask HN: Happy 404 Day. Whats your favorite 404 error page?

Top comment by HPsquared

Financial Times has my favourite: https://www.ft.com/404

A range of spurious economic explanations for why they couldn't serve the page.

4. Ask HN: Is anybody getting value from AI Agents? How so?

Top comment by JCM9

I’ve seen a lot of attempts but nothing that worked really well. Using an agent as a glorified search engine can work, but trying to replace actual humans to handle anything but the most standard use cases is still incredibly hard. There’s a lot of overhyped rhetoric at the moment around this tech, and looks like we’re heading into another period of post-hype disillusionment.

Legal angles here also also super interesting. There’s a growing body of scenarios where companies are held accountable for the goofs of their AI “assistants.” Thus we’re likely heading for some comical train wrecks as companies that don’t properly vet this stuff set themselves up for some expensive disasters (eg think the AI assistant doing things that will get the company into trouble).

I’m bullish on the tech, but bearish on the ability of folks to deploy it at scale without making a big expensive mess.

5. Ask HN: What is the current (Apr. 2024) gold standard of running an LLM locally?

Top comment by deathmonger5000

I created a tool called Together Gift It because my family was sending an insane amount of group texts during the holidays. My favorite one was when someone included the gift recipient in the group text about what gift we were getting for the person.

Together Gift It solves the problem the way you’d think: with AI. Just kidding. It solves the problem by keeping everything in one place. No more group texts. There are wish lists and everything you’d want around that type of thing. There is also AI.

https://www.togethergiftit.com/

6. Ask HN: Has Anyone Trained a personal LLM using their personal notes?

Top comment by appstorelottery

I've been disciplined (perhaps obsessive at times) with keeping a daily diary for many years and I was interested in being able to query my diary locally via AI. I found a solution that works surprisingly well using GPT4ALL.

I found GPT4ALL (https://gpt4all.io) to have a nice-enough GUI, and it runs reasonably quickly on my M1 MacBook Air with 8Gb of ram, and it can be setup to be a completely local solution - not sending your data to the Goliaths.

GPT4ALL has an option to access local documents, via the Sbert text embedding model (RAG).

My specific results have been as follows; using the Nous Hermes 2 Mistral DPO and Sbert - I indexed 153 days of my daily writing (most days I write between 2 and 3 thousand words).

Asking a simple question like "what are the challenges faced by the author?" provides remarkable, almost spooky results (which I won't share here) - which in my opinion are spot-on regarding my own challenges over that the period - and Sbert provides references to which documents it used to generate the answer. Options are available to reference an arbitrary number of documents, however the default is 10. Ideally I'd like to have it reference all 153 documents in the query - I'm not sure if it's a ram or a token issue, however increasing the value of documents referenced has resulted in machine lock-ups.

Anyhow - that's my experience - hope it's helpful to someone.

7. Ask HN: Freelance website builders/maintainers, what's in your 2024 toolkit?

Top comment by emsalazar

Depending on the clients' complexity and needs, I like to structure the work as:

Tier 1 (just need a basic site): Use the built-in website builders from their domain provider or hosted static site generators, like GoDaddy, Github Pages, etc.

Tier 2 (need plugins or ecommerce): If they require some more functionality like email lists or selling products then use services like Wordpress, Squarespace, or Shopify to build their site.

Tier 3 (extra data complexity): For sites that have more structured data that is not satisfied with tier 2, then use something like Webflow or Framer. These allow you to add more data types and categories to build unique pages.

Tier 4 (custom app) If they are building something unique that is complicated to build with the previous tiers, then my preferred stack to use is Django, Django Rest Framework, HTMX, Tailwind. Django-cms or Wagtail are also proven CMS to help build their site.

You should also consider serverless apps like Google Cloud Run or AWS Fargate so you don't worry too much about the infrastructure unless it is absolutely necessary to roll your own.

Tier 5 (a little bit of everything): Combine tiers with subdomains. For example, build their blog or marketing site on webflow, but have a custom API or an authenticated interface with a Django/DRF backend.

Ultimately, you need to see what level of technology your client is comfortable with and how you want to leave the project as your contract ends. Don't try to build a overtly complicated system that will give you more headaches later when the client calls on you again.

8. Ask HN: Going from CTO to Developer?

Top comment by simonw

I've done this. I have friends who have done this.

I'm personally a big fan of the "Engineer/Manager Pendulum" model described by Charity Majors: https://charity.wtf/2017/05/11/the-engineer-manager-pendulum...

Engineers who have spent some time in management are extremely valuable. They have a better grasp on how things work, they have better leadership skills (you don't need to be in a management role to provide useful leadership within a company) and they're generally useful as all-round contributors.

There will be people out there who will mark you down for this. That reflects poorly on those people, but that doesn't mean they won't be in positions of influence where their poor judgement can still impact you negatively.

There will be plenty of other people who get it, and who will respect the choices you have made.

And you can always swing the pendulum back again later.

9. Ask HN: Most efficient way to fine-tune an LLM in 2024?

Top comment by dhouston

Qlora + axolotl + good foundation model (llama/mistral/etc, usually instruction fine tuned) + runpod works great.

A single A100 or H100 with 80GB VRAM can fine tune 70B open models (and obviously scaling out to many nodes/GPUs is faster, or can use much cheaper GPUs for fine tuning smaller models.)

The localllama Reddit sub at https://www.reddit.com/r/LocalLLaMA/ is also an awesome community for the GPU poor :)

10. Ask HN: How to secure website for public launch

Top comment by gerardnico

The biggest security hole is user input.

Just escape every input: For sql, to avoid sql injection: https://datacadamia.com/data/type/relation/sql/parameter For html, if somebody try to inject html: https://datacadamia.com/web/html/entity

You got 99% of security holes patched.

All the best