Joined September 2008
- Tweets 2,060
- Following 245
- Followers 1,227
- Likes 1,023
518 Photos and videos
Gabor Herget retweeted
Jun 8
My favorite Claude design hack right now:
Create a prototype.md skill.
Instead of one-shotting UI, I have Claude generate 5 variations for every feature in a single HTML file, side-by-side.
I pick the winner. Claude builds it.
74
91
2,501
165,638
FSD approved in Denmark
FSD Supervised now approved in Denmark 🇩🇰
Rollout will begin soon
5,816
10,775
96,376
23,476,734
Gabor Herget retweeted
Jun 2
You can now train 120B parameter models locally on a laptop! 🔥
We collabed with NVIDIA and Microsoft to bring LLM training on the 128GB unified memory RTX Spark laptop!
This is the NVIDIA RTX Spark Superchip. A new beginning for personal computers.
Designed for creators, AI developers, and gamers, RTX Spark brings over 30 years of NVIDIA innovation to slim Windows laptops and small, ultra-efficient desktop PCs.
55
105
1,134
89,322
Gabor Herget retweeted
May 28
build the thing that builds the thing.
openclaw.ai/ecosystem
65
59
785
55,372
May 29
The brutal first years when you start your own business
I started many, but never got easier
IMO if you are good at smth, owning it is the only option
May 28
My salary as an employee:
Year 1: $54K (Big 4 accounting)
Year 2: $60K (worked at startup)
Year 3: $80K (startup)
Year 4: $105K (software engineer)
Year 5: $115K (software engineer)
Then, I quit my job to build Starter Story.
My revenue as an owner:
Year 1: $0
Year 2: $12K
Year 3: $67K
Year 4: $186K
Year 5: $497K
Year 6: $752K
Year 7: $1.6M
Year 8: Sold the company.
Bet on yourself.
1
119
Introducing Claude Opus 4.8: it builds on Opus 4.7 with sharper judgment, more honesty about its own progress, and the ability to work independently for longer than its predecessors.
Available today at the same price.
ALT Benchmark table showing how Claude Opus 4.8 compares to its predecessor and to other models on tests of coding, agentic skills, reasoning, and practical knowledge work tasks.
3,684
8,622
67,441
15,253,321
May 28
AMD is 0.76% of MSCI World Index right now, so my indirect holding is just a bit less than what I was directly holding. Still painful, but could be worse...
May 28
Wanted to hold my AMD for 4 years, then sold too early 🤦
At least, I still own AMD indirectly through my MSCI World Index ETF...
Do what I say, but don't do what I do: gherget.com/investing-guide
56
May 28
Wanted to hold my AMD for 4 years, then sold too early 🤦
At least, I still own AMD indirectly through my MSCI World Index ETF...
Do what I say, but don't do what I do: gherget.com/investing-guide
90
May 28
Just running my fully local AI company using Paperclip and my DGX with MiMo-V2.5-coder-Q2
Quite slow at 55 tok/s but at least 1,993 tok/s for prompt processing (using unsloth studio)
Paperclip is using the local pi agent hooked up with this model on my DGX
huggingface.co/jedisct1/MiMo…
May 28
When openclaw came out, I started building an AI dev team with UI/UX designers, two devs, and QA, using hacks like local state files and GitHub issues.
It worked quite well but I was missing an overview, and claw was not quite right for this. It stopped at random times, and I had to add new models all the time.
I thought about building something myself, but then paperclip came out, and it does exactly did this.
It's really cool, and stable. I can use my local agents like claude code, hermes, pi and plug them in. I started with a CEO and let him build his AI team. I am basically the board member who guides the CEO and team members with decisions and goals!
It comes with issues, an inbox, and an API that I even gave to my claude code if I wanted to cleanup issues, tasks or files.
131
May 28
When openclaw came out, I started building an AI dev team with UI/UX designers, two devs, and QA, using hacks like local state files and GitHub issues.
It worked quite well but I was missing an overview, and claw was not quite right for this. It stopped at random times, and I had to add new models all the time.
I thought about building something myself, but then paperclip came out, and it does exactly did this.
It's really cool, and stable. I can use my local agents like claude code, hermes, pi and plug them in. I started with a CEO and let him build his AI team. I am basically the board member who guides the CEO and team members with decisions and goals!
It comes with issues, an inbox, and an API that I even gave to my claude code if I wanted to cleanup issues, tasks or files.
173
May 28
My work setup in the last few years:
2019 vscode (typing code with my hands!)
2022 cursor ai (typing prompts, and reviewing code)
2026 openclaw, hermes, claude code (reviewing plans, letting the agents create online accounts, and juggling UIs for me)
today: dropping IDEs completely and using paperclip, and cmux. Building a team of ai agents who review each other, using cmux to delegate work to codex and claude code sessions
1
78
Gabor Herget retweeted
May 27
Introducing local Reachy Mini conversations: free chats forever!
So fast that we had to hardcode delays to stop it from interrupting you mid-sentence.
We built an open-source Realtime API powered by llama.cpp:
Parakeet -> Gemma 4 E4B -> Qwen3TTS
Run it anywhere you run local LMs. Video shows DGX Spark and a 36GB M3 Pro MacBook.
Blog: huggingface.co/blog/local-re…
40
62
412
75,822
Gabor Herget retweeted
May 24
Q: How are job postings for software engineers rising rapidly despite AI agents automating coding?
A: Because there’s far more code to manage than ever before. We’re already seeing a 14x YoY increase in GitHub commits, and it’s accelerating.
AI has dramatically lowered the cost of writing code, so it’s now being used across far more businesses, applications, and use cases.
We’re at the beginning of a massive productivity boom driven by the proliferation of bespoke software throughout the entire economy.
Coding has been AI’s breakout use case this year. The fact that it’s increased demand for software engineers — rather than decreased it — should call into question the entire “AI will cause mass job loss” narrative.
827
1,466
9,116
1,631,242
Gabor Herget retweeted
May 23
I'm late to the party, but cmux is great. github.com/manaflow-ai/cmux
current split:
codex mac app: knowledege work, learning, reading
cmux codex cli: coding
282
224
3,786
527,043
