Hacker News: serced

New comment by serced in "Why can't transformers learn multiplication?"

serced — Fri, 24 Oct 2025 18:36:25 +0000

Yes, I also wonder about this! Progress from children books to scientific papers etc. Could it learn e.g. language structure faster in a pre-training stage? Also somehow one needs to define a proxy to generalization to compute a loss and do backpropagation.

New comment by serced in "Why do LLMs freak out over the seahorse emoji?"

serced — Mon, 06 Oct 2025 06:47:28 +0000

I find interesting that their blog post on prompt/context engineering kind of stands against their ultra long system prompt. Maybe it is not too specific as in their visual example (too specific - just right - too vague). https://www.anthropic.com/engineering/effective-context-engi... and the system prompt https://docs.claude.com/en/release-notes/system-prompts#sept...

New comment by serced in "I'm absolutely right"

serced — Fri, 05 Sep 2025 14:18:11 +0000

It's nice to see Claude.md! I checked out the commits to see which files you wrote in which order (readme/claude) to learn how to use Claude Code. Can you share something on that?

New comment by serced in "Apertus: An open, transparent, multilingual language model"

serced — Tue, 02 Sep 2025 09:20:36 +0000

Great to see this from a public institution here! Can you share evals and the tech report? On Huggingface it leads to a 404 for me.

New comment by serced in "Polars Cloud: The Distributed Cloud Architecture to Run Polars Anywhere"

serced — Fri, 07 Mar 2025 22:51:51 +0000

May I ask what part in M&A needs this much data processing? I am quite familiar with the field but did not yet see such tasks.

New comment by serced in "Software engineering job openings hit five-year low?"

serced — Fri, 21 Feb 2025 10:15:31 +0000

Zurich IT market died after COVID. Not sure about other hubs.

New comment by serced in "Ask HN: What Are You Working On? (October 2024)"

serced — Mon, 28 Oct 2024 12:53:01 +0000

To get some experience launching webapps that can be put into the play store and play around with image generation models, prompting, I am building an app to generate application/corporate photos from a few non-professional selfies.

Tech stack is probably FastAPI (I mainly know python) and likely nuxt/ionic (none/not much experience). Not sure how the whole hosting, interaction with replicate/huggingface will work on phone apps, payments on stripe without having a company, how to make the webapp into a phone app, etc. It should be a great learning project with the first time scoring an actual sale! Happy to hear early guidance if people have done similar things with python backgrounds to get started.

New comment by serced in "Show HN: I made a git rebase TUI editor"

serced — Mon, 14 Oct 2024 16:42:00 +0000

TIL there is 'vimtutor'. I barely new the basics for quickly creating a file and inserting stuff. Will have a look if there is something to learn there, thanks for the pointer.

New comment by serced in "Ask HN: Does anyone use sound effects in their dev environment?"

serced — Mon, 16 Sep 2024 19:26:03 +0000

In the previous startup I worked at, we set up a PR action that played a celebration thingy after every PR merge. We used https://github.com/leokster/dingdong_sonos to play on the Sonos speakers. Fun little gimmick but not really as a usecase like colors for linting etc.

New comment by serced in "Understand and train your own GPT – No prior AI or coding knowledge"

serced — Thu, 30 May 2024 14:10:46 +0000

Hi all I wrote this "blog" Google colab notebook for friends and family who have little technical know-how or do not yet have an understanding of language models. It's based on nanoGPT (a minimal GPT implementation from Andrej Karpathy, ex-OpenAI - not sure whether there is a need to state this here). However, I rewrote the code from scratch, made it more explicit, and have lots of textual information such that everyone should be able to follow what is happening when, and train their own little GPT :) Hope it helps some people and let me know if you have any questions/feedback.

Tl;dr: Blog about how LLMs work & train a mini model without prior knowledge.

Understand and train your own GPT – No prior AI or coding knowledge

serced — Thu, 30 May 2024 14:10:46 +0000

Article URL: https://colab.research.google.com/drive/1J9qJ-JArJSORVfOcppoD4cMC3u2I7aGZ?usp=sharing

Comments URL: https://news.ycombinator.com/item?id=40523938

Points: 4

# Comments: 1

New comment by serced in "Taking Risk"

serced — Wed, 22 May 2024 11:17:10 +0000

Just from a qualitative perspective, i.e., personal experience/discussions.

New comment by serced in "Taking Risk"

serced — Wed, 22 May 2024 06:48:54 +0000

Interesting but also scary to go this route. I think part of the problem in Europe is also that companies/governments are less open to just try out new (startup) solutions. I would love to be learn from US startups how they manage to sell without much done already.

New comment by serced in "Show HN: Share events to your friends' calendars"

serced — Mon, 06 May 2024 18:31:14 +0000

Good idea, thanks! Generally we may need to add a short paragraph that no data is stored as this may not be clear.

Show HN: Share events to your friends' calendars

serced — Mon, 06 May 2024 15:59:51 +0000

This is about a cool hack for a trivial problem a friend and I faced. We couldn't invite friends to a summer party because we didn't have some of their email addresses. Next best option? Share a text in a Whatsapp group. However, does everyone keep the event in mind when planning other things?

Hence, we came up with save2cal.com. This javascript app allows you to create your event and share the link with your friends. When the link is clicked, it opens the calendar app with your content prefilled such that they can directly save the event. Happy to hear your thoughts!

One thing we'd love to change is making the links shorter. We don't store anything on the server side and encode all entered information in the string and decode it in the browser when the link is clicked. Then we generate the calendar file and open it. Ideas to make the link shorter? We could create a shortlink mapping, but then we would need to store data. We don't want that.

An example: https://save2cal.com/v1/index.html?data=JTdCJTIydCUyMiUzQSUy...

Comments URL: https://news.ycombinator.com/item?id=40276123

Points: 2

# Comments: 2

New comment by serced in "Ask HN: Who wants to be hired? (May 2024)"

serced — Thu, 02 May 2024 14:52:38 +0000

Location: Switzerland / Europe

Remote: Yes

Willing to relocate: Potentially SF? Otherwise preferably hybrid/remote, though I am open to exploring options.

Technologies: Machine Learning (CV, NLP, etc.), Software Enigneering (FastAPI, Pydantic, SQL, a bit of frontend & IaC using Terraform with Microsoft Azure).

Résumé/CV: https://www.linkedin.com/in/severinhusmann/ or happy to send CV via message/mail.

Email: seevgame@hotmail.com

Unique background combining machine learning studies/work experience (ML/software engineer at Swiss startup) with business studies/experience (consulting, mergers and acquisitions, private equity, startup). At the startup, I worked on a computer vision project and as a software engineer on building their latest SaaS product (FastAPI, Pydantic, SQLAlchemy, some frontend in Vue.js, etc). I would love to work with cool people building the future using AI :)