Hacker News: greggh

New comment by greggh in "I don't recommend Tailwind CSS"

greggh — Sun, 02 Aug 2026 12:26:28 +0000

Many of us oldies who started working on the web years before IE 6 actually like Tailwind. We all have our preferences. But I know multiple 50+ year olds who have been doing this work since CSS was barely a proposal who now use Tailwind.

New comment by greggh in "Advancing the price-performance frontier with GPT‑5.6"

greggh — Thu, 30 Jul 2026 23:54:18 +0000

Use a harness like OMP that lets you choose which model does which things. My main model is GLM 5.2, it handles planning and anything I dont have covered by other models. Tasks from todos and in sub agents are done by deepseek, I have different models for the git work like add/commit/push (that goes through cheap Minimax M3), and so on...

This way the expensive/strong model only handles the architecture and orchestration tasks. The cheaper models handle everything else and the strong one knows how to tell them what to do in enough detail to get good work out of them.

New comment by greggh in "Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac"

greggh — Wed, 29 Jul 2026 17:25:08 +0000

Yeah, its the 256gb version.

New comment by greggh in "Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac"

greggh — Wed, 29 Jul 2026 15:59:56 +0000

It does exactly what it says it does. On my Mac mini M4 with 16GB of ram it is running at just over 5 tok/s. That jump from M4 to M5 is crazy.

New comment by greggh in "Show HN: Orate – On-device neural text-to-speech queue for Mac"

greggh — Tue, 28 Jul 2026 15:28:49 +0000

It took the key! I tried emailing the support email in the website footer and it bounced back. Thought it might be easier to work through things.

It still crashes when I try to view the shortcuts in settings. But everything else seems better.

New comment by greggh in "Show HN: Orate – On-device neural text-to-speech queue for Mac"

greggh — Fri, 24 Jul 2026 00:34:14 +0000

It doesn't crash on me anymore! But it won't take my license key, and I am just hitting the copy button on polar.sh to copy the key and paste it in.

New comment by greggh in "Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA"

greggh — Fri, 24 Jul 2026 00:27:08 +0000

Stock OMP. When you add a provider and then /model you can choose a model and choose what roles it will take. Do that for each role it has.

Then I use /plan when I need to be planning and not writing to any files. Take the advice the UI gives you where it tells you to add the word orchestrate into your prompts when you want to make sure it uses its todo/tasks and sub-agents.

New comment by greggh in "Show HN: Orate – On-device neural text-to-speech queue for Mac"

greggh — Wed, 22 Jul 2026 21:52:17 +0000

I had to click Skip to get past the intro screens, then the app comes up. I went to settings and pasted in my license, clicking activate says it's not valid. It is the key on my polar.sh account and I copy/pasted it so I know it's correct. Then it crashed.

New comment by greggh in "Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA"

greggh — Wed, 22 Jul 2026 21:43:43 +0000

I do this in OMP, a fork of Pi. It lets you set different models for different tasks. So with an API that has many different companies models I can set the Plan model to the best one, right now I am using GLM 5.2 for that, it plans really well. I have Vision set to Kimi 2.7 Code (cheaper and vision is just fine). Minimax M3 is set to the Advisor role (double checks work). Deepseek v4 Flash is set for the Task role. And MiMo 2.5 pro is set as default.

With this setup GLM handles planning and managing my AGENTS.md, and orchestrating subagents for tasks from the plan/todo GLM created. The tasks themselves are handed off to Deepseek v4 flash to implement with strong instructions and examples for each agent. Minimax M3 reviews the output as the Advisor and recommends changes, catches bugs, and whatnot, subagents can be re-run with that information.

Overall I am saving a lot using some of these smaller models. But with this setup I am getting great results.

New comment by greggh in "Show HN: Orate – On-device neural text-to-speech queue for Mac"

greggh — Wed, 22 Jul 2026 00:39:03 +0000

Mac mini m4 16gb, latest updates.

New comment by greggh in "Show HN: Orate – On-device neural text-to-speech queue for Mac"

greggh — Tue, 21 Jul 2026 22:40:21 +0000

It would be nice to have a sample of the audio quality and voice(s) it uses on the website so we know what it will sound like.

-- Edit:

I installed it, ran it and got the first time wizard, granted permissions and hit continue and it just exits without any other message or screen. It isn't in the menu bar. I run it again and get the same. It says the permissions are granted already the second time though.

I like the interface on the website a lot, so I would love to use this. Thoughts?

New comment by greggh in "Where are YC founders now? OpenAI and Anthropic, mostly"

greggh — Thu, 16 Jul 2026 13:11:20 +0000

The wheels on the bus go round and round, round and round, round and round...

New comment by greggh in "Grok Build is open source"

greggh — Wed, 15 Jul 2026 22:34:03 +0000

Just a prototype? I have no reason to leave the terminal for a GUI IDE. TUI works great, does what I need and is very easy to use and interact with.

New comment by greggh in "1024000^2 Blocks, 2B2T Minecraft Server World Download Project, and Discoveries"

greggh — Tue, 19 May 2026 13:33:47 +0000

2b2t Place is exactly that.

https://2b2t.place

New comment by greggh in "This Month in Ladybird – April 2026"

greggh — Sun, 03 May 2026 14:10:10 +0000

Uhh, yes? Non-profits take donations to keep doing their work.

New comment by greggh in "TIL: Apple Broke Time Machine Again on Tahoe"

greggh — Sun, 01 Feb 2026 22:53:09 +0000

On Tahoe my Time Machine was broken after the update. My backup target is on a QNAP NAS. I just had to set it up from scratch again and it worked. But it did cost me a few files I was trying to recover. So I feel this.

New comment by greggh in "Ross Stevens Donates $100M to Pay Every US Olympian and Paralympian $200k"

greggh — Thu, 29 Jan 2026 02:11:59 +0000

The real answer here is that he is mad about people protesting what Israel is doing in Gaza. This $100M donation is being made with funds he had given to UPenn. He has taken it back, via lawyers, because they allowed the protests to go on. He is now just taking that original donation and moving it somewhere else. Not that I am against the Olympians getting paid, just some context.

Sources: https://philanthropynewsdigest.org/news/donor-pulls-100-mill... https://thehill.com/homenews/education/4348656-upenn-loses-1... https://www.timesnownews.com/world/who-is-ross-stevens-stone... (many more)

New comment by greggh in "Trinity large: An open 400B sparse MoE model"

greggh — Wed, 28 Jan 2026 21:54:16 +0000

The only thing I question is the use of Maverick in their comparison charts. That's like comparing a pile of rocks to an LLM.

New comment by greggh in "Claude Cowork runs Linux VM via Apple virtualization framework"

greggh — Thu, 15 Jan 2026 19:14:43 +0000

Use a devcontainer. Claude Code's repo has one built specifically for it:

https://github.com/anthropics/claude-code/tree/main/.devcont...

New comment by greggh in "Stop Doom Scrolling, Start Doom Coding: Build via the terminal from your phone"

greggh — Wed, 07 Jan 2026 02:01:32 +0000

Following that story as it happened, it was all on the phone with the phone keyboard and he somehow made multiple good Neovim plugins including that very popular one (which I use in multiple configs).