Arthur @arthurliebhardt.com · 23/09/2026Can someone explain how this is possible, that medium effort performs better than xhigh? Is it overthinking / overengineering for dev tasks? 000
Arthur @arthurliebhardt.com · 18/09/2026agree, jev or jev like models will have a huge impact it's not the " shiny" new thing but rather really useful and fast high-volume decision making tool go to x and just search jev for the diffent things people build already - harness decisions like 100
Arthur @arthurliebhardt.com · 21/08/2026expected another boring, badly designed dev site got absolutely surprised echarts.apache.org/en/index.html 000
Arthur @arthurliebhardt.com · 20/08/202610h generation with /goal burned 65% of the $200 max plan result was solid but was it worth it? 🙃 would've changed some UI decisions if I just prompted it better along the way probably a skill issue on my end with /goal 000
Arthur @arthurliebhardt.com · 19/08/2026Grok Bot is now my IDE @cursor_ai + @bot is an awesome combo I build with Cursor, hand off tasks to Grok Bot, which can spin up Cursor again if needed Both have a mobile app I can check on the go. 100
Arthur @arthurliebhardt.com · 14/08/2026I tried using Fable as the orchestrator and Grok 4.6 High as the executor in @cursor_ai The result was terrible, the UI it generated was so bad. Does anyone have a good workflow for this? look at this login form 🤣 000
Arthur @arthurliebhardt.com · 14/08/2026back on @cursor_ai + Cloud Agents finally not fighting storage issues every day 512 gb mac was a mistake, never again 010
Arthur @arthurliebhardt.com · 13/08/2026grok 4.6 tend to generate such hero sections somehow I like it 😁 000
Arthur @arthurliebhardt.com · 04/08/2026will share more about enterprise and B2B AI here creating a lot of enablement materials for SAP Business Technology Platform (soon Business AI Platform) and maybe some of you might find it useful you get a real feeling where enterprise AI is at right now 100
Arthur @arthurliebhardt.com · 02/08/2026get a VPS -> spin up codex -> let it deploy n8n and browser-use -> automate whatever you want 000
Arthur @arthurliebhardt.com · 02/08/2026Show codex / claude a skill or let them use your computer and make a reusable skill out of it rerun it as often you want 🙂 000
Arthur @arthurliebhardt.com · 23/07/2026GPT 5.6 Sol med - xhigh is currently my go-to coding model been testing fable 5 and GPT 5.6 Sol the last few days Opus 4.8 with dynamic workflows was already pretty good but fable tops this. GPT 5.6 Sol Ultra also spins up a lot of subagents but the results weren't that good in my tests 210
Arthur @arthurliebhardt.com · 29/06/2026I’ve been building a small agentic GitOps infrastructure on k3s, and the main lesson is this: the agent should not be the platform. GitOps is the platform. The agent is an operator that works inside boundaries. Current shape of the stack: 1. k3s on Hetzner 100
Arthur @arthurliebhardt.com · 23/06/2026I do a lot of work around agent-assisted coding/vibe coding, and two questions come up every single time: How do I get my app live? How do I make it secure? my first instinct was an agentic cloud that just auto-checks and ships it 320
Arthur @arthurliebhardt.com · 15/04/2026TIL you can type ! in claude code to execute shell commands directly and then keep working with the results previously I just opened another tab and copy pasted everything like back and forth 120
Arthur @arthurliebhardt.com · 13/04/2026@cursor_ai cloud agents are wild you literally spin up a VM, tell it what to build, and watch it: - open browsers - click around - run tests - record everything - debug itself first agent walks you through setup and gets everything running without you lifting a finger 100
Arthur @arthurliebhardt.com · 11/04/2026what are your experiences with @convex? How fast did you hit the limits of the free tier and what is possible with the 25 dollar plan? 000
Arthur @arthurliebhardt.com · 10/04/2026I introduced some bugs with the latests changes and focusing on fixing them today-md v1.7.6. 100
Arthur @arthurliebhardt.com · 10/04/2026told hermes my openai-codex subscription wasn't working hermes, say no more ... 🤦♂️ 000
Arthur @arthurliebhardt.com · 06/04/2026looks like they just waited until I bought my Mac Studio so I could go all in on local models let's see how Gemma 4 performs 011
Arthur @arthurliebhardt.com · 01/04/2026I'm not sure if I fell into the local agent trap, but I'm very bullish on local models and autonomous agent frameworks like OpenClaw or Hermes Agent Let's see what an M4 Max is capable of 100
Arthur @arthurliebhardt.com · 28/03/2026TIL what Hand off actually does 😅 a few days ago I was stuck in a PR, wanted to start a parallel feature but had no clue how to create a worktree Hand off does exactly this Git worktrees let you check out multiple branches at once in separate directories. all linked to the same repo. 100
Arthur @arthurliebhardt.com · 27/03/2026tried SwiftUI Plugins to give today-md that Liquid Glass look results were... absolutely terrible 🤣 sticking with the old design 120
Arthur @arthurliebhardt.com · 27/03/2026Sync in your Obsidian Vault and open / edit / plan you tasks there in markdown plan your tasks in today-md 🫡 100
Arthur @arthurliebhardt.com · 27/03/2026I’ve been building my own small cloud setup on Hetzner instead of using managed hosting for everything. The setup is pretty simple: 100
Arthur @arthurliebhardt.com · 26/03/2026today-md v1.7.3 release is out Right now, calendar support works through the native Calendar app on Mac, which means connected iCloud, Google, Outlook, and Exchange calendars show up there first. This release tightens up the calendar workflow: 121
Arthur @arthurliebhardt.com · 25/03/2026shipped first version of the calendar planning feature the challenge wasn't the logic but getting UX and UI right, thinking through the full flow, edge cases, UI behavior behind every move auto adjusts lanes when tasks move between Today, This Week, Next Week 210
Arthur @arthurliebhardt.com · 24/03/2026building the next productivity tool everyone types, then lets AI fix grammar I just want to speak AI transcribes and corrects in real-time = more output, same time of course native mac app with local LLMs for security and compliance 400
Arthur @arthurliebhardt.com · 23/03/2026everyday I recognize a small improvement I need for my day to day work My calendar is pretty packed that I often need to block time for deep work / specific tasks. Fantastical had such feature but was blocked on my work laptop. let's ship it! 000
Arthur @arthurliebhardt.com · 23/03/2026shipped a landing page for today-md using Lovable the workflow is simple: export a PRD from codex, paste it into Lovable, tell it to build a landing page, make some tweaks based on taste done in maybe 30 min todaymd.app/ (company VPN could block this) 000
Arthur @arthurliebhardt.com · 23/03/2026Some improvements to access and create tasks easier - Global Dynamic Island for creating tasks - MenuBar to see today's tasks, mark them as done, and create new ones - Light / Dark / System mode selection in the settings 100
Arthur @arthurliebhardt.com · 20/03/2026Another Day, Another Open Source Project recall. A local-first macOS app for turning long audio and video into searchable transcripts you can chat with on-device. 240
Arthur @arthurliebhardt.com · 20/03/2026today-md v1.5.8 Not a single line in this app was touched by me. All Codex, a little bit of Claude, and some taste/preferences. 100
Arthur @arthurliebhardt.com · 19/03/2026switched to OpenRouter as the model provider instead of my Claude subscription cause OpenClaw burns through tokens like crazy GPT 5.4 keeps writing whole poems, useful findings, but stops tasks in between Now testing Xiaomi's MiMo-V2-Pro, curious what it can handle 010
Arthur @arthurliebhardt.com · 18/03/2026improved it a bit based on my workflow - Better search with highlighted matches - Search previews for notes and checklist content - Faster Markdown editing with shortcuts - Keyboard shortcuts cheatsheet - Better multi-select task actions - Bulk mark tasks as done - Drag tasks straight into Done 210
Arthur @arthurliebhardt.com · 16/03/2026OpenClaw is pretty nice but burns through tokens like crazy hoping local LLMs catch up with the latest models for daily tasks soon need to try Qwen 3.5 9B anyone using it daily? 000
Arthur @arthurliebhardt.com · 16/03/2026if you want to check which local models will run on your device 👇 www.canirun.ai/canirun.aiCanIRun.ai — Can your machine run AI models?Detect your hardware and find out which AI models you can run locally. GPU, CPU, and RAM analysis in your browser. 000
Arthur @arthurliebhardt.com · 16/03/2026While the main projects generate, I vibe-coded a small macOS task app called today-md. It’s built with SwiftUI and keeps planning simple: Today, This Week, and Backlog, plus Markdown notes, subtasks, drag-and-drop, and local JSON backups. local-first, fast, and intentionally lightweight. 210
Arthur @arthurliebhardt.com · 15/03/2026OpenClaw 🤝 marketing finally took some time to automate stuff I was doing manually for ai-media-studio marketing images now I have a 24/7 marketing companion that just runs my exact workflow 110
Arthur @arthurliebhardt.com · 14/03/2026moved all my projects from Vercel to Hetzner Over the last week, I migrated everything to a small k3s setup. Main reason? cost and flexibility to self host what I need. 110
Arthur @arthurliebhardt.com · 28/02/2026any todo app / work system recommendations? tried everything but with context switching between multiple private projects / 9-5 / private life etc I always end up in a mess 😅 010
Arthur @arthurliebhardt.com · 14/02/2026Claude Cowork is underrated. It thinks more clearly and delivers better solutions. 000
Arthur @arthurliebhardt.com · 11/02/2026pov: building a product in Germany all my side hustle profit is gone for tax account invoice roughly 2k every year 🥲 i tried doing it with claude cowork, and it gave me the exact same numbers as the tax consultant, already in an excel sheet, ready to enter and send to the german tax office 000
Arthur @arthurliebhardt.com · 08/02/2026Better late than sorry. If you don't have a good guide or tutorial, setup could take some time. Thankfully, people in @tinkererclub shared the perfect video. have some ideas I want to try out 100