Hacker News: cocoflunchy

Made by Mechanical Means

cocoflunchy — Mon, 28 Sep 2026 07:06:06 +0000

Article URL: https://felixrieseberg.com/made-by-mechanical-means/

Comments URL: https://news.ycombinator.com/item?id=49874551

Points: 61

# Comments: 23

New comment by cocoflunchy in "How I changed teaching after AI managed to do all my homework assignments"

cocoflunchy — Sun, 27 Sep 2026 13:40:02 +0000

I wonder if you could see this the other way round? Prior to AI you were limited in your ability to grade handwritten tests.

But now you could have the students take a written exam every other week and use a LLM to grade it?

I'm pretty sure Opus 5.5 has good enough vision capabilities to auto grade with the right prompt.

New comment by cocoflunchy in "Roboharm: Do frontier robot policies refuse unsafe instructions?"

cocoflunchy — Mon, 21 Sep 2026 21:22:19 +0000

Is this a useful benchmark if the doll is obviously non-human? Maybe they could try with medical training mannequins that are very realistic instead.

Fractal basins trap latent reasoning

cocoflunchy — Tue, 08 Sep 2026 19:07:38 +0000

Article URL: https://arxiv.org/abs/2609.04963

Comments URL: https://news.ycombinator.com/item?id=49615194

Points: 1

# Comments: 0

New comment by cocoflunchy in "Building an interactive instrument for a one-of-a-kind festival"

cocoflunchy — Wed, 02 Sep 2026 10:27:45 +0000

Kind of like "oh-clare". It means clear waters.

New comment by cocoflunchy in "Launch HN: Bloomy (YC S26) – AI-powered mastery learning for K-12"

cocoflunchy — Mon, 20 Jul 2026 20:59:36 +0000

I think 383toast is talking about your landing page design which has very strong AI-generated design vibes (blue-green gradients, rounded rectangles with a colored border on the left...) There are a bunch of skills you can find online to improve on this if you want to keep designing with Claude, and the latest generation of models are better (Fable/GPT5.6).

Check https://github.com/jakubkrehel/skills or https://github.com/emilkowalski/skills (more around animations), maybe https://github.com/leonxlnx/taste-skill (didn't try it)

All the evidence-based advice we found on how to be more successful in any job

cocoflunchy — Tue, 07 Jul 2026 11:51:46 +0000

Article URL: https://80000hours.substack.com/p/all-the-evidence-based-advice-we

Comments URL: https://news.ycombinator.com/item?id=48816444

Points: 3

# Comments: 0

New comment by cocoflunchy in "Ask HN: Who is hiring? (July 2026)"

cocoflunchy — Thu, 02 Jul 2026 04:41:27 +0000

Dashdoc is a SaaS Transport Management System for road freight. We help carriers and shippers run transport operations on one shared platform.

We’re hiring a product-minded engineer to help us expand in Belgium and the Netherlands. Local specifics: e-CMR mandates, digital waste regulations, Dutch/Flemish accounting standards, regional carrier workflows often block deals or deployments. Your job is to find those blockers and turn them into shipped product.

You’ll work directly with customers, sales, CSMs, and product. You’ll be able to spend time with carriers in Belgium and the Netherlands, watch users use the product, read contracts/regulations, then spec, build, deploy, and measure the features yourself.

Stack: React, React Native, Python/Django, Django REST Framework, PostgreSQL, Google Cloud. We use Shape Up, ship continuously, and use AI coding tools like Claude Code, Cursor, and Codex.

Good fit if you speak Dutch fluently, write clearly in English, like owning ambiguous product problems end-to-end, care about simple UX and clean code, and enjoy deleting code as much as adding it.

Apply: https://dashdoc.welcomekit.co/companies/dashdoc/jobs/product...

New comment by cocoflunchy in "Qualcomm to Acquire Modular"

cocoflunchy — Wed, 24 Jun 2026 20:58:11 +0000

It's the first sentence of the article? "an all-stock deal valued at nearly $4 billion"

New comment by cocoflunchy in "DeepSeek Introduces Vision"

cocoflunchy — Thu, 18 Jun 2026 21:44:35 +0000

I may have misremembered but I thought I had read somewhere that recent models by OpenAI and Anthropic tend to produce reasoning that is not always understandable for humans. But you're right that it's not the case for Deepseek so maybe I'm hallucinating ;)

Or maybe it was an article or a tweet about researchers trying really hard to steer the model to think in English otherwise interpretability / safety becomes a lot harder?

New comment by cocoflunchy in "DeepSeek Introduces Vision"

cocoflunchy — Thu, 18 Jun 2026 08:46:51 +0000

I believe that most reasoning models actually think in their own "language" which is not really understandable by humans. The thinking traces that are shown in the UI are actually summaries generated by a smaller model in plain english (or user language). Sometimes this leaks through and you see some chinese/japanese characters in e.g. Claude's reasoning.

New comment by cocoflunchy in "Five frontier LLMs disagree on 67% of 1k real-world fact-check claims"

cocoflunchy — Thu, 28 May 2026 14:10:18 +0000

It's not in the training data, so there is no way for the model to know.

New comment by cocoflunchy in "HashiCorp co-founder says GitHub 'no longer a place for serious work'"

cocoflunchy — Wed, 29 Apr 2026 13:52:12 +0000

Mainly this https://mrshu.github.io/github-statuses/ But also product decisions that don't seem to be aligned with the preferences of users... see https://mitchellh.com/writing/ghostty-leaving-github

New comment by cocoflunchy in "Launch HN: Twill.ai (YC S25) – Delegate to cloud agents, get back PRs"

cocoflunchy — Sat, 11 Apr 2026 02:14:32 +0000

Sent you some feedback from the app, I can't get GitHub to connect. Feel free to contact me over email to troubleshoot!

New comment by cocoflunchy in "Launch HN: Twill.ai (YC S25) – Delegate to cloud agents, get back PRs"

cocoflunchy — Sat, 11 Apr 2026 02:02:46 +0000

Great timing as I'm exploring the space to get rid of Cursor in our stack. For local dev everyone is switching to Claude Code or Codex. The state of the art for cloud agents in my opinion right now is Cursor. But their pricing model per-user doesn't make sense when what I want is to enable anyone in the company to fix things in the product. 2 things not immediately clear from your homepage: - do you support full computer use? Again Cursor is the best I've tried there - what kind of triggers do you support? We have in particular one automation built with cursor to auto approve PRs that are low-risk. It triggers on a specific comment on a PR Finally some advice from a user's pov: you need to invest a lot in the onboarding experience. I tried Devin today and it couldn't get it to work after one hour of fiddling. How do you store the repo's setup scripts? Cursor cloud is pretty opaque and annoying to configure on that side. Anyway I'll try it!

New comment by cocoflunchy in "Apple Business"

cocoflunchy — Tue, 24 Mar 2026 17:15:07 +0000

I'm currently in that hellish process too... I don't know how to get out of it. Did you know that your employees will be forbidden from downloading from the App store once you launched that migration? It's a nightmare

New comment by cocoflunchy in "Ask HN: Who is hiring? (February 2026)"

cocoflunchy — Mon, 02 Feb 2026 17:05:14 +0000

Dashdoc | France | REMOTE (from France) | Full-time | https://www.dashdoc.com/

Dashdoc is a fast-growing transport management platform for the road freight transportation industry.

We're hiring a Lead Software Engineer to work on our Invoicing domain. Small 3-person (including you) team, great impact on tens of thousands of people using the tool all day everyday, great colleagues, international context, access to good tools (Cursor, Claude Code).

https://dashdoc.welcomekit.co/companies/dashdoc/jobs/lead-so...

New comment by cocoflunchy in "Ratchets in software development (2021)"

cocoflunchy — Mon, 02 Feb 2026 11:30:24 +0000

Also see https://www.notion.com/fr/blog/how-we-evolved-our-code-notio...

New comment by cocoflunchy in "France Aiming to Replace Zoom, Google Meet, Microsoft Teams, etc."

cocoflunchy — Mon, 26 Jan 2026 17:26:03 +0000

Visio from https://lasuite.numerique.gouv.fr/

New comment by cocoflunchy in "The Code-Only Agent"

cocoflunchy — Mon, 19 Jan 2026 13:53:28 +0000

Why would it recreate sed and awk? The screenshot from the repo even shows it using sed.