Hacker News Reader

lpellison 60% Fable cost cut by converting code to images and having the model OCR it

I tried the same thing last year (with openai models), back then it worked to reduce prompt tokens, but you needed way more completion tokens, ultimately more expensive (and slower) https://pagewatch.ai/blog/post/llm-text-as-image-tokens/

lpellison Ask HN: What Are You Working On? (April 2026)

Still working https://pagewatch.ai/ , my ai readiness audit tool. Currently having fun building a MCP app for it inside Claude / ChatGPT, its oddly tricky to get things behaving consistently.

lpellison Launch HN: Sentrial (YC W26) – Catch AI agent failures before your users do

The github link is also going to a 404.

I built a tool to check for these issues, was curious if it would find it all, but yes.

https://pagewatch.ai/s-bm6jq1qs6y1x/b560hmfx/dashboard/previ...

lpellison Ask HN: What Are You Working On? (March 2026)

Ironically I'm getting SSL handshake failed on your site

lpellison Ask HN: What Are You Working On? (March 2026)

Really pretty keyboard widget, though it did slightly confuse me that is showing the next key, not what I actually typed.

lpellison Ask HN: What Are You Working On? (March 2026)

There is so much opportunity in AI that is not just a chatbot, I almost feel there should be category of tools that is LLM powered, but not [here is empty textbox] Best of luck!

lpellison Ask HN: What Are You Working On? (March 2026)

https://pagewatch.ai/ - make sure LLM's can understand your site.

There is a surprising amount of edge cases that can cause ChatGPT or others to misunderstand your pages. Some models can handle div based tables, some want alt tags but cannot understand title tags, etc.

I built the tool to check your site as close as possible to what a human would see and then compare it with LLM's.

It was a weird journey trying to tease this info out of the models, they will happily lie, skip checking sites or just make things up.

lpellison An Update on Heroku

I've been very happy with https://render.com/ , seems the closest to what heroku was

lpellison How to code Claude Code in 200 lines of code

I think this might be one of the first times I didnt notice it, but just look through the comment history of https://news.ycombinator.com/threads?id=jackfranklyn , they all look the same.

lpellison Ask HN: What are you working on?

https://pagewatch.ai/

Building a tool to check your site layout and copy from multiple devices. Uses gpt-5 vision to find inconsistencies in headings/images.

lpellison Show HN: I made simple Email Marketing app with base features

Landing page is pretty but some serious spelling errors going on in the main title ('Flixible Email', 'Bussines'), might want to fix that :) see https://pagewatch.ai/s-jbjiz2qxvndc/li4tigvz/dashboard

lpellison Show HN: I made a landing page audit checklist

FIY your homepage automatically loads new content when you scroll down, making things look quite chaotic and impossible to see the footer.

lpellison Rivian's TM-B electric bike

This thing at least looks like a bike, some of the 'e-bikes' are way more motorcycles than actually meant to be paddled.

lpellison Django 6.0 beta 1 released

A task framework could be very useful, setting up celery task can get complicated very fast, especially once you have multiple servers with rolling deploys or recurring task (celery-beat). Rails has active-job and is always something I felt was missing from Django.

lpellison Replacing a $3000/mo Heroku bill with a $55/mo server

https://render.com/ is probably the closest, I'm really enjoying using them. Workflow is the same as heroku, but cheaper, no nightly restarts, supports new python versions etc..

lpellison The scariest "user support" email I've received

Pretty clever to host the malware on a sites.google.com domain, makes it look way more trustworthy. Google should probably stop allowing people to add content under that address.

lpellison Webflow Donates $150k to Support Astro's Open Source Mission

Astro is really fun to work with, cool that you can add components from vue/react etc all on the same page. Sounds like they are planning to use the money for their AI builder, hopefully the core keeps getting worked on also.

lpellison Ask HN: Why buy domains and 301 redirect them to me?

Do you have an affiliate plan, or likely to have one? Maybe they plan to redirect with their affiliate ID at some point?

lpellison Show HN: Get your website copy and design roasted

Ha not bad, pagewatch.ai

  Look at that video player, apologizing profusely because it doesn’t know why it 
  can’t load. Sadly, it’s the most relatable thing on your webpage. Your site is 
  like a desperate dating profile: "I swear I'm different!

I wanted to reciprocate with my tool (not as spicy) https://pagewatch.ai/5f5c1cef-c119-4389-9641-965a4fee2b33/29...

lpellison UI elements with a hand-drawn, sketchy look

It is cloudflare that blocked us, we scrape using a regular chrome instance (with the latest user-agent), but cloudflare are sometimes very aggressive in their blocking. I'm busy adding a proxy option to handle these cases automatically, and some way to whitelist our scraper.

lpellison UI elements with a hand-drawn, sketchy look

I'm using this for the AI annotations at https://pagewatch.ai/ Its strange how much different it makes the whole page feel from regular boxes, much more of a tangible/informal feeling. Felt perfect to go with the type of feedback LLM's give.

lpellison Launch HN: Nango (YC W23) – Source-available unified API

I'm struggling a bit to figure out the pricing here. If I have lets say 50 users that sync Trello, does that count as 50 synced connections?

lpellison GPT-4 Turbo with Vision is a step backwards for coding

Same thing happened to me, I'm in South Africa

lpellison Show HN: I scraped 200M Shopify products to build a search engine

Seems like a fun scraping project, I think you have to work on extracting more accurate categories though, for example this link does not really include snowboards for me: https://www.searchagora.com/search?query=Snowboard And the first products I clicked have rather weird descriptions, https://www.searchagora.com/products/snowboard-bd2a90aa-6808...

Maybe its my location (South Africa) but I also cannot visit the product store when I click through

lpellison Stract: Open-souce, non-profit search engine

Seems surprising ok for coding related queries ('celery rate limit'), I'm curious about their scraping setup, building that out must be quite a big task.

lpellison Show HN: SingleAPI – Convert the Internet into your own API

This is pretty cool, it is able to parse data out of a random pricing table somewhere in the page. It does seem to just make up data it if is not found in the page (probably expected with LLM's), I wonder if you can reduce that with some prompting, or maybe verify the data is actually present? Your schema page docs is broken https://singleapi.co/docs/schema

lpellison Show HN: Phind V2 – A GPT-4 agent that’s connected to the internet and your code

phind is awesome, a few times it has helped me where GPT-4 could not due to its knowledge cutoff. some minor feedback on the UI, it doesn't always scroll to the bottom if it is generating code, and I have to manually scroll it into view

lpellison Persistence of Vision Test

Surprising how well I can see the image even after putting the separation to 64. The image is just a few thin strips yet I can easily see the whole thing if I follow the scrolling, really cool.