HNHacker News
TopNewBestAskShowJobs

mihau

2,582 karma · joined November 29, 2014

https://github.com/myhau
submissionscomments
mihau··on Show HN: Cactus Needle 3: 8-29MB automation models can match DeepSeek V4 Flash
None of the queries I asked worked:

- "more light"

- "less light"

- "both doors should be locked"

- "if blinds are open, open back door"

mihau··on I vibed a proof of Conway's conjecture
Really nice "proof guide": https://gaearon.github.io/conway-refinement/#/highlights

and "proof map": https://gaearon.github.io/conway-refinement/#/map/conway-ref...

mihau··on The Problem
what was the latest model you tried?
mihau··on MiniMax H3 Day-0 Support in ComfyUI: Open Weights, Native Audio, and 2K Video
Has anyone tried running this on Mac device? (e.g. Mac Studio Ultra)
mihau··on Streaming directly the K3 official Hugging Face 1.6TB of weights on m5 max 128gb
https://xcancel.com/antirez/status/2082136334160818528
mihau··on Claude Opus 5
30% on ARC-AGI-3
mihau··on Window Hints – Keyboard-first window switching
Hello!

I’m a big fan of tools and apps that make common (preferably all) actions directly accessible from the keyboard. For example, I use Vimium every day, and it makes me so much faster. When I'm in the zone, using it feels like I'm "flying" :D

At the same time, people often use multiple displays with tons of windows open across multiple apps. Switching between these using the mouse, cmd+tab or Spotlight (by typing the app's name) kinda works, but it's easy to lose your flow when doing so.

That's why I built Window Hints. It lets you jump directly to any window using only your keyboard.

To get a feel for how it works, it's best to watch some demos on the website:

https://windowhints.com

- When you press a keyboard shortcut, Window Hints shows labels over all visible windows. Type a label to jump to the corresponding window. You can also press the shortcut again to include buried windows.

- It also works when you reveal all windows with Mission Control (ctrl+↑ or swipe up with 3 fingers), as well as when Stage Manager is enabled.

- Additionally, some apps and tools allow you to create splits. Window Hints currently supports them in Chrome, Ghostty, and tmux. More integrations to come.

The app is native and weighs ~10MB. While it's already quite fast, I'm going to keep trying to squeeze every last bit of performance out of it.

(There are some challenges on macOS, though, especially with certain apps that can be slow to respond even when using just cmd+tab.)

I want to continue working on this while making it somewhat sustainable, so there is also a paid version: a lifetime license for $7. Feature set stays the same, but every 50th use in the free version shows a mildly annoying pop-up that you can immediately dismiss. That's it.

Feedback is incredibly important to me at this stage, so I'd really appreciate your thoughts and ideas!

I will read every comment in this thread, but you can also submit feedback here: https://windowhints.com/feedback

mihau··on I'm 33 and I think Claude Code is melting my brain
https://nitter.net/BraedendotTECH/status/2077353000486547633
mihau··on Ask HN: What Are You Working On? (July 2026)
Yes, if you have 3 Chrome windows, it will show you visual hints for all of them separately
mihau··on Ask HN: What Are You Working On? (July 2026)
Working on Window Hints - keyboard-first window switching (macOS for now)

https://windowhints.com/

mihau··on Anthropic officially bans using subscription auth for third party use
I agree that Claude Code is buggy as hell, but:

> Let's not talk about the feature parity with other agents.

What do you mean feature parity with other agents? It seems to me that other CLI agents are quite far from Claude Code in this regard.

mihau··on Lyria 3
same
mihau··on Gemini 3
@simonw wen pelican
mihau··on Craft Chrome Devtools Protocol (CDP) commands with new command editor
FYI it's behind a feature flag (aka experiment). Just in case you rarely use experiments in DevTools:

DevTools -> Settings (cog, top right) -> Experiments -> Search for "Protocol Monitor" -> Check the checkbox

mihau··on Apple M5 chip
why? M3 Ultra already had 800 GB/s (6400 gbps) memory bandwidth
mihau··on Why the push for Agentic when models can barely follow a simple instruction?
It feels to me that the OP on the forum expects this to work: "read this existing function, then read my mind and do stuff" (probably followed by "do better").

It still takes a lot of practice to get good at prompting, though.

mihau··on Two things LLM coding agents are still bad at
> you can overengineer your prompt to try get them to ask more questions

why overengineer? it's super simple

I just do this for 60% of my prompts: "{long description of the feature}, please ask 10 questions before writing any code"

mihau··on The case for an iceberg-native database
Kinda off-topic but I love the quality of images used in this article. Even memes are HD. Quite rare as people tend to use whatever meme creator provides - super-pixelated, low-quality source image with text slapped on top.
mihau··on Interactive Double Pendulum Playground
Yep, it's just bad. Physics feels totally off.
mihau··on China bans tech companies from buying Nvidia's AI chips
https://archive.is/lJuT0
mihau··on Homomorphically Encrypting CRDTs
Sorry for going off-topic, but kudos for UI/UX on your blog!

To name a few: Nice style (colors, font, style), "footnotes" visible on the margin, always on table of contents, interactivity and link previews on hover.

Nice. What's your tech stack?

mihau··on Show HN: My LLM CLI tool can run tools now, from Python code or plugins
Thanks for feedback!

Ah, I saw "llm.Toolbox" but I thought it's just for plugin developer convenience.

I'll take a look at the issue you posted (#1111). Maybe I can contribute somehow :).

mihau··on Show HN: My LLM CLI tool can run tools now, from Python code or plugins
Hi Simon!

I'm a heavy user of the llm tool, so as soon as I saw your post, I started tinkering with MCP.

I’ve just published an alpha version that works with stdio-based MCP servers (tested with @modelcontextprotocol/server-filesystem) - https://github.com/Virtuslab/llm-tools-mcp. Very early stage, so please make sure to use with --ta option (Manually approve every tool execution).

The code is still messy and there are a couple of TODOs in the README.md, but I plan to work on it full-time until the end of the week.

Some questions:

Where do you think mcp.json should be stored? Also, it might be a bit inconvenient to specify tools one by one with -T. Do you think adding a --all-tools flag or supporting glob patterns like -T name-prefix* in llm would be a good idea?

mihau··on Did ChatGPT just message me first?
Link to ChatGPT conversation: https://chatgpt.com/share/66e718e5-c934-8001-8ffa-ef7ca1f165...
mihau··on A disk so full, it couldn't be restored
Be careful with iPhones/iOS too. Ignoring „not enough space” warnings will eventually put it into a boot loop.

There are threads on Reddit about this.

For me, flashing it through Finder with the official firmware was the only option. I lost some photos, the rest I was able to restore from iCloud backup.

mihau··on macOS Sonoma 14.4 might break Java on your machine
Does Apple consider this to be a serious issue? Does anyone know if Apple plans to release a fix soon, perhaps in version 14.4.1?
mihau··on Wing: A cloud-oriented programming language – request alpha access
If anyone is interested in Infrastructure as Software, I also recommend giving Pulumi a try. I've been using Terraform for a couple of years, Pulumi for about a year and I think that Pulumi is on a completely different level.

It also supports higher level abstractions (Component Resources). For example: check out AWS EKS provider.

mihau··on What the “superforecasters” predict for major events in 2023
Do these "superforecasters" lose something when they are wrong? Do they have "skin in the game"?

I'm a big fan of predication markets (e.g. Polymarket, PredictIt) for exactly that reason – proper incentives are there.

mihau··on The PS5 Has Been Jailbroken
This is already kinda-released: https://twitter.com/SpecterDev/status/1576714840650883074
mihau··on Brave Search Goggles: Alter search rankings with rules and filters
Nice! Ability to quickly choose goggle would be nice, something similar to "site:" or !bangs in ddg. Example search: "integration tests best practices goggle:tech-blogs".
Page 1 of 2Next →