zapnuk

At scale this make a difference. For me/us not that much.

At my work we currently don't have a subscription and use API based billing.

Some time ago the price for my usage was about 30-50€ per day when we used opus for about everything. With luna+sol its about 10€ for sol to plan or debug and 5€ for luna to implement.

At that point it doesn't quite matter to me if the cheap anthropic model is 20-50% more expensive compared to luna since luna is so unbelievable cheap. The big gamechanger was the cheaper models being good enough to do the implementation given a good "rough" plan.

Nice that that's also possible with anthropics models.

The things stated in the announcement?

Mainly node compatibility though more test coverage and different performance improvements.

Actually HN handles it almost perfectly. The automatic line breaks are really good.

Only complain would be that its sometimes difficult to see the parent comment(s) when you scroll down.

HN could fix this quite easily by making the current parents sticky while you scroll.

There is a reasone we use left and right margin/padding.

This blog is quite unreadable for 27/32" monitors.

We should be greatful for this. This is the one public case study on how large-scale llm-driven code generation actually works out.

With node and deno there are reasonable alternatives for everyone who don't want to use bun anymore.

You could also use the LLM create a program to do the conversion and then review and use the program to deterministicly perform the actual conversion.

Have the best of both worlds.

GitLab? We use gitlab for work. Its way worse in comparison.

Last week I encountered a bug where my merge request simply didn't show that I deleted a file. Apparently it's because my MR included the creation of a folder with the same name as the basename of the deleted file. Unacceptable for a code hosting platform.

Other than that I miss GH Actions, a clear ui (gitlab has way too many sub-menus), a responsive ui (gitlab feels very sluggish). And while we don't have the Gitlab duo activated, it still pops out regularly eventhough I can't use it besides closing it. ...and I don't even want to start with their issue buard.

It strongly reminds me of Jira in terms of quality, which is no compliment.

US models align with our "average" (western) values. If we outsource thinking by using LLMs, why would we outsource it to an LLM that doesn't have our values encoded in it?

Couldn't be more clear violation from a legal standpoint.

Though its quite sad that the community had more creativity (and engineering talent) to develop classic(+) wow.

Everything Blizzard now touches is bland, lacks soul, or is straight up bad.

Sure, a poweruser can bring their own ram/ssd. But again they pay almost as much and have a worse system performance wise.

Normal users don't profit from anything you listed. They do have to buy a notebook with all components, and thus currently have to pay more for linux/windows hardware compared to Apple.

Also, RAM isn't backwards compatiple. Literally had this problem with my old ddr4 not fitting in the newer ddr5 slots when my ddr5 acted up.

Framework Laptop is more expensive than a Macbook Air with all around worse hardware. For a framework 13 I'd have to pay 1900€ with a 16GB setup. For 1450 I get a MBA with 24GB ram. Similar with a dell or lenovo who get smoked in performance comparisons.

It might still be worth it for those who hugely value open source and repairability but as for value I think its save to say that Apple is currently in a league of their own. Even if the altest os update is a flop.

Also, the Macbook has improved repairability. While its still not great its better than a few years ago.

Thats just one of the interpretations of a skill.

A skill can also act as an abstraction layer over many tools (implemented as an mcp server) to save context tokens.

Skills offer a short description of their use and thus occupy only a few hundled tokens in the context compared to thousends of tokens if all tools would be in the context.

When the LLM decides that the skill is usefull we can dynamically load the skills tools into the context (using a `load_skill` meta-tool).

Gemini 3 was:

1. unreliable in GH copilot. Lots of 500 and 4XX errors. Unusable in the first 2 months

2. not available in vertex ai (europe). We have requirements regarding data residency. Funny enough anthropic is on point with releasing their models to vertex ai. We already use opus and sonnet 4.6.

I hope google gets their stuff together and understands that not everyone wants/can use their global endpoint. We'd like to try their models.

Thats like being proud of not using google or stackoverflow and only reading manuals, or using notepad instead of an IDE (or editor with language server support).

A 10$ GitHub Copilot or 20$ ChatGPT/Claude subscription get you a long way.

And if the employer isn't willing to spend this little money to improve their workers productivity they're pretty dumb.

There are valid concerns like privacy and oss licences. But lack of value or gain in productivity isn't one of them.