Andreas Hegenberg

4.7K posts

Andreas Hegenberg

Andreas Hegenberg

@LLo_ai

🦥 Developer of BetterTouchTool and BetterSnapTool. Founder of https://t.co/tLFeyWYHE1 GmbH (however this is my private Twitter account). Mastodon: @[email protected]

Munich Katılım Eylül 2009
571 Takip Edilen5.2K Takipçiler
Andreas Hegenberg retweetledi
Yigit Konur
Yigit Konur@yigitkonur·
nobody's talking about BetterTouchTool, but they just jumped into the AI agent game with this new "ask AI" feature that builds macOS automations for you. you can even plug in codex / claude code / opencode etc, which is lowkey insane for a first beta. if you're an automation nerd who lives on your mac, this unlocks a ton of workflows. don't let the UI throw you off, yeah it's a little rough around the edges, but don't sleep on it. personally i've built 100+ workflows over the years using BTT. at this point i use macOS almost entirely on keyboard and in a fully AI native way, all made possible by the little patch-like workflows he's created. it's genuinely changed how i use my computer. the dev behind BTT (@LLo_ai) is straight up one of the most talented macOS engineers i've ever come across (big shoutout to @stephancasas too). dude's been putting in serious work for 10+ years with no VC money, no big backing, just him quietly building one of the most underrated hidden gems in the entire macOS automation ecosystem. built and maintained largely on his own. that's rare. that's real. honestly the potential here is massive. this is exactly the kind of thing VCs should have on their radar. he could genuinely fix a lot of what's broken in the macOS AI ecosystem (no thanks to apple for years of shipping mid). if he goes all in on this, the upside is basically unlimited. would love to see @OpenAI scoop this up and plug it into the @OpenAIDevs ecosystem. it could be something really special. cc: @sama and yeah, ideally @apple would be the perfect acquirer here, this could make millions of mac users' lives genuinely better. but let's be real, apple is too big to notice the things quietly doing the most important work. unfortunately that's just how it goes. whoever picks this up is getting a serious deal. big props to @LLo_ai, hope this is finally the moment he gets the success he's always deserved. rooting for you till the end. 🙏
Yigit Konur tweet media
English
2
8
23
1.2K
Andreas Hegenberg retweetledi
Wes Bos
Wes Bos@wesbos·
@theCTO I recently discovered you can use better touch tool by @LLo_ai to totally replace logi options. It’s excellent software
English
7
1
90
12K
Andreas Hegenberg retweetledi
juan
juan@juanbuis·
to commemorate alan dye moving from apple to meta, here's one of his best quotes
juan tweet media
English
417
4.1K
65.3K
2.1M
Andreas Hegenberg
Andreas Hegenberg@LLo_ai·
BetterTouchTool's Black Friday sale just went live! 25% off standard licenses, 35% off lifetime licenses! folivora.ai/buy
English
3
2
8
1.2K
Andreas Hegenberg
Andreas Hegenberg@LLo_ai·
@wesbos I also updated the status item management features in BTT. docs.folivora.ai/docs/6e_manage… It's still pretty crazy - they reworked the whole architecture for Tahoe but you still can't access hidden items without third party tools that use very hacky ways to access them
English
0
1
2
183
Wes Bos
Wes Bos@wesbos·
Any good Bartender alternatives? Bartender went wacky with the Tahoe update. I used Vanilla for a few years, but I need re-order, and quick search for menu bar items. Raycast needs to save us here!
English
25
3
87
81.5K
Andreas Hegenberg
Andreas Hegenberg@LLo_ai·
It feels like it should be supported because of how the new Spotlight was promoted. It would allow apps to add a lot of interactivity to it. However can this really be completely broken on Tahoe RC? SnippetIntents work when triggered from the Shortcuts app.
English
0
0
0
877
Andreas Hegenberg
Andreas Hegenberg@LLo_ai·
I didn't find much information on this. Does anybody know whether macOS Tahoe supports SnippetIntent in Spotlight? I tried Apple's AppIntentsTravelTracking example, but at least there it doesn't seem to show the interactive view when triggered via Spotlight on Tahoe RC.
English
2
0
0
1.2K
Andreas Hegenberg
Andreas Hegenberg@LLo_ai·
@steipete I also don't think the models got worse. The only thing that many of us experienced was degraded performance - which made it slower to use and maybe contributed to that psychological effect.
English
0
1
0
1.6K
Peter Steinberger 🦞
Peter Steinberger 🦞@steipete·
I don’t see claude (code) or other models getting worse, I believe it’s more a psychological effect where we just keep expecting more and the initial wow-effect is gone.
English
42
9
204
17K
Andreas Hegenberg
Andreas Hegenberg@LLo_ai·
@steipete There’s currently a hard split into two camps of developers. 1.) “It’s nice for simple things, but you absolutely can’t generate complex code with it. I’ve tried Copilot before, and it didn’t help me much.” 2.) “I haven’t written a single line of code by hand in two months.”
English
1
1
0
142
Peter Steinberger 🦞
Peter Steinberger 🦞@steipete·
Kinda nuts how many companies are not yet on the agent train. They’ll be steamrolled before they know it.
English
14
3
60
9.8K
Andreas Hegenberg
Andreas Hegenberg@LLo_ai·
Definitely adding support for Spotlight App Actions to BetterTouchTool, I have been waiting for an extensible Spotlight for quite a while. Let’s see how powerful these app actions are.
English
1
1
16
968
Gary Thompson ☮
Gary Thompson ☮@gwthompson·
@LLo_ai This is extremely exciting! The possibilities are endless! I use a cli tool called LLM from @simonw that has jrecently gained support for using MCPs via a plugin! The integration of LLM, BTT and MCP has my brain exploding with ideas! 🤯
English
1
0
0
73
Andreas Hegenberg
Andreas Hegenberg@LLo_ai·
AI integration in BetterTouchTool, beta will start sometime next week: #t=0" target="_blank" rel="nofollow noopener">vimeo.com/1091121762?sha… Will require your own API keys, can use most cloud AI services but also local LLMs via e.g. LMStudio. Follow community.folivora.ai/t/h-llo-ai-upc… for updates
English
3
3
8
1.2K
Nitin
Nitin@gniting·
@pertikaer Thanks for the response. I've already tried what you have suggested. However, once I wake up the system from sleep, the Raycast assigned hyper key stops working and the setup process needs to be redone. I am happy to be a beta tester if you need me to test anything.
English
2
0
2
199
Andreas Hegenberg
Andreas Hegenberg@LLo_ai·
@joyless_urgency I'm working to improve the condition window, it will be better soon. Currently you'd need to use a dynamic variable defined in JS and reference that in your condition:
Andreas Hegenberg tweet mediaAndreas Hegenberg tweet media
English
1
0
2
69
Daniel Hilse
Daniel Hilse@joyless_urgency·
Also nested variables here would be really helpful. I'm finding myself wishing I could add a condition like "If focused_window_x < (focused_screen_x/2)"
English
1
0
0
35
Daniel Hilse
Daniel Hilse@joyless_urgency·
@LLo_ai First of all love BTT, long time power user. Just starting to lean into the possibilities of conditions and wow, I'm getting a tong of ideas. It would speed up the process though if my selected parameters could highlight or optionally filter the "Current State" readout!
Daniel Hilse tweet media
English
1
0
0
48
Andreas Hegenberg
Andreas Hegenberg@LLo_ai·
BetterTouchTool's Black Friday sale is now live: 25% off on standard, 33% off on lifetime licenses! folivora.ai/buy
English
3
7
15
6.9K
Andreas Hegenberg
Andreas Hegenberg@LLo_ai·
@film_girl I have been using Qwen2.5-Coder-32B-Instruct (MLX 4bit) today and it has been really good at coding tasks. Got ~23 token per seconds with small contexts and ~15t/s on larger contexts. I assume this also works well on M3 Max.
English
0
0
0
64
Christina Warren
Christina Warren@film_girl·
@LLo_ai Llama 3.1 (3.2 would be interesting too since it’s smaller) would be great to see but pick your fave!
English
2
0
0
134
Andreas Hegenberg
Andreas Hegenberg@LLo_ai·
The only benchmark that matters to me: Clean build of BetterTouchTool on M1 Max: 182s, on M4 Max: 99s (in general building seems significantly slower on Xcode 16 than on older Xcode versions )
English
5
4
37
10.3K
Andreas Hegenberg
Andreas Hegenberg@LLo_ai·
@film_girl So I'm just doing some very first tests with LM Studio, should probably first read a bit into the whole world of local LLMs :-) It's already working pretty well. For small context sizes I'm getting around 9 t/s with this model: Meta-Llama-3.1-70B-Instruct-GGUF feels really usable
English
0
0
1
115
Christina Warren
Christina Warren@film_girl·
@LLo_ai Really curious to see your LLM benchmarks to compare against my M3 Max. I’m not upgrading but I want to know how much FOMO I should have.
English
1
0
0
184