bennett_dev

  Location: Germany  
  Remote: Yes  
  Willing to relocate: Yes, Europe preferred
  Technologies: TypeScript, React, Next.js, Python, PostgreSQL, AWS, Terraform, Docker, AI agents / MCP  
  Résumé/CV: https://hollstein.io/cv.pdf
  Email: [email protected]
  GitHub: https://github.com/vantezzen

Fullstack/product engineer, mostly working with TypeScript and Python. I'm especially interested in developer tools, AI/agent infrastructure and small teams with a lot of ownership.

I maintain AutoForm, an open-source TypeScript library with 3.5k+ GitHub stars and 50k+ weekly npm downloads. I also built a TikTok Year in Review product that handled ~10M visits in one month, and currently work on an AI agent product using TypeScript/Python, MCP, subagents and human-in-the-loop workflows.

I feel these Claude-written blog articles always have such a weird speech to them, not that fun to read

From CloudFlare's side (https://blog.cloudflare.com/deno-joins-cloudflare/) is seems celld was the main point of this - letting the Deno team focus on making workerd self-hosting easier.

I feel it leaves a bitter flavor how Ryan Dahl pushed so hard for Deno and Deno Deploy for years, just to let them die within 1 year and 6 months respectively. Thankfully I don't have any codebases that heavily use Deno features, otherwise this would be a steep curve now.

  Location: Germany  
  Remote: Yes  
  Willing to relocate: Yes, Europe preferred
  Technologies: TypeScript, React, Next.js, Python, PostgreSQL, AWS, Terraform, Docker, AI agents / MCP  
  Résumé/CV: https://hollstein.io/cv.pdf
  Email: [email protected]
  GitHub: https://github.com/vantezzen

Fullstack/product engineer, mostly working with TypeScript and Python. I'm especially interested in developer tools, AI/agent infrastructure and small teams with a lot of ownership.

I maintain AutoForm, an open-source TypeScript library with 3.5k+ GitHub stars and 50k+ weekly npm downloads. I also built a TikTok Year in Review product that handled ~10M visits in one month, and currently work on an AI agent product using TypeScript/Python, MCP, subagents and human-in-the-loop workflows.

There’s a separate „Who wants to be hired“ thread for this

Especially with „ sales conversion probability“ it just doesn’t sound universal to other issues - there’s tons of unique models for specific use cases

Interesting that the Under Attack Mode wasn’t used at all here. I understand not wanting to break APIs but I feel temporarily challenging non-API usage could have at least helped without impacting users too much?

Next add realistic signal delay so you have to wait a day to get the result

AFAIK Ladybird is separated from SerenityOS completely now (https://hackaday.com/2024/07/02/fork-ladybird-browser-and-se...) so the Ladybird contributors can fully focus on it

Sounds similar to Vercel‘s portless CLI (https://portless.sh/)

There was a nice task on 39c3 https://youtu.be/3OAiOfCcYFM

They do try to upstream and eventually just have Linux natively support Apple Silicon!

I'm not familiar with Elysia but how does this handle subfunctions? e.g. if my code is

const app = new Elysia() .patch('/user/:id', (request) => { return handleUserRequest(request); })

or some other custom logic, does it automatically need to fall back to full parsing?

Such an interesting showcase of history. RCT was a masterpiece written in Assembly from hand. Looking at code from OpenRCT2 you can really see the thought behind it.

Now ISO Coaster really is a badly slipped together adaptation. A lot of details don’t make sense, graphics often look weird and performance is terrible - all while using tons of compute resources.

Same - most of the time I just directly open the specific status (e.g https://http.cat/504 )

Nothing to do with politics and everything to do with crappy AI slop. There was some list somewhere of some of those reports and it was painful to look at

What differentiates this from http://fingerprint.com/?

Visited for the first time and it said I already visited 800+ times with a 99.5% accurancy - not very promising. From the code this also looks like very simple client-side fingerprinting + IP information?

Exactly my thoughts. The whole time scrolling I thought „and now a full example? Ah no, another block of extreme details I dont need“. If it’s so easy and great - just show me what it looks like

This plus ChatGPT scheduled tasks could make an interesting alarm clock