Torn

Hmm no there's no little evidence to support the body getting used to external melatonin and not making its own in response

On the plus side, LLM slop like this in readmes are a great indicator that you can just close the tab and carry on with life - nothing of value lost

+1. I think there's an interesting idea here but an AI slop writeup on an AI slop website is very tiring

I think it's a bit different in this case. There are plenty of modern synthesisers, from Korg and others

This is intended to be a faithful recreation of a classic synth, for collectors / rich enthusiasts

This is awful, but why is this on hacker news

> Now an international team, led by Flinders University, have found that in a small number of people, the immune system can accidentally confuse a normal adenovirus protein with a human blood protein termed platelet factor 4 (or PF4).

Seems to have been a legitimate, very rare, side effect

https://www.flinders.edu.au/research/articles/covid-vaccine-...

This is a really great watch from Benn Jordan on the topic https://www.youtube.com/watch?v=_bP80DEAbuo

Some wild things happening with those, and infrasound. Colossus is shown 4 mins in

Canvas is not Canva. Is this a bot reply

Interesting - Claude immediately refuses

     API Error: Claude Code is unable to respond to this request, which appears
     to violate our Usage Policy (https://www.anthropic.com/legal/aup). Please
     double press esc to edit your last message or start a new session for
     Claude Code to assist with a different task. If you are seeing this refusal
     epeatedly, try running /model claude-sonnet-4-20250514 to switch models.

Hopefully not a security incident

Tbh I find self-documenting CLIs (e.g. with a `--help` flag, and printing correct usage examples when LLMs make things up) plus a skill that's auto invoked to be pretty reliable. CLIs can do OAuth dances too just fine.

MCP's remaining moats I think are:

- No-install product integrations (just paste in mcp config into app)

- Non-developer end users / no shell needed (no terminal)

- Multi-tenant auth (many users, dynamic OAuth)

- Security sandboxing (restrict what agents can do), credential sandboxing (agents never see secrets)

- Compliance/audit (structured logs, schema enforcement)?

If you're a developer building for developers though, CLI seems to be a clear winner right

https://github.com/microsoft/playwright-cli and https://agent-browser.dev/

Mostly, yes: https://github.com/microsoft/playwright-cli

I personally found playwright-cli, and agent-browser which wraps playwright, both more token-efficient than using the raw mcp.

Odd that this article from Dec 2025 has been posted to the top of HN though

Disney 100% has access to colorists and best in class colour grading software. It must have been a business (cost cutting) decision?

Could your processing work in a webview? i.e. webgl or webasm or similar and you communicate with it via postMessage. Something like Polygen might help with the scaffolding, but I have not tried it personally

You'd think spotify as a mature company would have had obligations to report this stuff!

I've had good experience contacting canva support in the past who've checked things with engineering teams, you should be able to contact them about their new affinity sign-in

Is enshittification truly inevitable? Perhaps, but as an occasional affinity user this is good news and I don't see a reason to spell doom just yet

I imagine enough folks will pay for the Canva account subscription to upsell - and then it's also a funnel into Canva

Plus it directly attacks Adobe's moat if a solid desktop app competitor is free

Yeah as someone who uses Claude code daily this feels like hype and marketing

CLAUDE.md and working memory only goes so far — it never religiously follows them and does not truly 'learn' from past collaboration. In fact the more you put into a CLAUDE.md the more it seems to degrade

Great writeup. There are other scaling axes of course, around data (even synthetic data) and improving AI generation at the 'software' layer (smarter design / training efficiencies / inference speed ups) — progress in those might make the the currently-unthinkable orders of magnitude $500b and beyond not as necessary?

We are 100% going to get 'hey remember when LLMs were pure and not explicitly (or more dangerously: subtly) recommending things' nostalgia in years to come.

There are parallels to early web here I'm sure of it.

I think I'm a little more worried about AI being subtly influenced in its training data -- they can't explain why they give the tokens they do, and even chain of thought / explain your working thinking is similarly made up and hallucination-prone

It to be a viable currency? Not a speculative asset?

Siri is nowhere near an AI assistant or even a decent AI query bot.

It mostly does simple commands and for the rest kicks you to web search results.

Apple surely understand the opportunity here, maybe their view is LLMs are just too variable / hallucinate too much to go all-in on it?

Yes, very lucrative with huge free cash flow. 60% YoY growth last year too

because ultra processed foods are objectively bad for us?

This is a great writeup and deep-dive into the conflicts of interest. Sounds like they are over-valued and simply _have_ to go into secondary markets

Looks like a project with 45k stars, where the latest update was Oct 2020. Why the HN post without a submission comment? Slick site I guess?

iframes also work pretty well for this kinda thing :D