Daigo

I live in Mountain View and we rent. The other day, someone asked me on LinkedIn what was the point not spending while earning high. We don't live in a mansion but our kids are happy. Instead of eating out expensive food, our kids love what my wife and I cook every day. We see a neighbor kid sent to school by a hired driver while we walk and talk to the elementary school every morning as a whole family. Instead of stressing out at a big tech, I run a one-person tech business on my term. My wife and I saved and invested enough and now our annual investment return alone is much bigger than the money we spend. So whether we work is a choice. We are in our healthy 40s and still love working. We aren't too thrifty but never upgraded our lifestyle like others do around here in Silicon Valley. That was the key for the growing assets. We don't look like rich people, but we have time freedom, choice, and peace of mind thanks to that. (Wow, that was a long reply.)

Tech employees are to be either laid off or subjected to incessant cognitive overload while babysitting fast-typing AI agents. Such dystopia. Own a business run by AI instead of working for AI.

Focus on originality and automate everything else.

I chuckle when the AI agent says “let me do it manually”

Qwen 3.6 is awesome

I played with Claude Design, starting by letting it read OmniCreator's frontend code. I hit the usage limit, but the menu interactions are pretty much complete and look polished! (...now I need to wait for a week.)

Why is openclaw tui so much better than web ui?

AI agents promise automation. But can you trust them to deliver, every time? Here’s what often gets overlooked: Most AI agents, as powerful as they are, still behave unpredictably. They make different decisions on each run—sometimes nailing the task, other times missing the mark. If your goal is consistent, reliable results, traditional software still matters. Hardcoded solutions run the same logic every time. No improvisation, no guesswork. The real unlock: AI agents calling robust, reusable software for the parts that demand precision and repeatability. This blend lets agents work faster and with fewer surprises. As agentic workflows spread, reliability hinges on having good, accessible software behind the scenes.

Good software saves tokens when used by AI agents.

I didn’t expect old leadership habits at a Y Combinator startup to resurface when I started working with AI agents: Moving fast isn’t always progress. At a YC startup, I watched sharp junior engineers run full speed. My job? Don’t slow them down. But here’s the catch: Speed without direction is wasted. If you skip business context or set the wrong foundation—API design, database schema, core dependencies—talented people just accelerate toward rework. I learned to step in early, only on the points that are hard to change later. The rest? Let the team run. Now, my only teammate is AI. The lesson’s the same: unchecked AI code makes you fast, but just as often, fast in the wrong direction. People complain about AI output. I see the tradeoff—because I’ve lived on both sides. What’s your approach when you’re the one setting guardrails—for humans or AI?

I'm setting up OmniCreator.club's proper agent permission system on the backend so I can use an agent to get LinkedIn post scheduling done from chat.

Work is becoming metawork.

You cannot lead without mastering delegation. I was terrible at delegating for my entire career. But something changed last week for an unexpected reason. It happened while I worked with the Home AI Agent. (My agent is a DELL Pro Max with DB running Nemoclaw with locally running LLMs.) I was making the main agent to delegate tasks to sub-agents. The main agent should remain responsive to my requests while the subs are busy-buzzing. With the context window limit and a telephone game between me, the main agent, and the subs, it was quite a bit of work to make a reliable protocol. The whole process became efficient and repeatable. A great success! It was still easier for me to make the delegation work because I didn't have to worry about their emotions. While my mind was focused on this task, I may have made delegation my new habit. I was even terrible at delegating at home. But I just delegated the weekend laundry duty to my 11-year-old daughter, and she did it 🎉 My new obsession with the home AI agent is already paying off 😁

AI agents work extremely inefficiently without software. (Just imagine they have to code for every task every time.)

It’s been four weeks since I sketched out the idea of giving instructions to an AI agent from the gym. I’ve now automated the entire process: planning SEO blog content, research, writing, generating header images, and publishing to the site. Goodbye to the Danish writer who had been doing this work until now. And I achieved all of this without sending any token fees or personal data to Anthropic or OpenAI. I even operated it from the gym right away. (Only image generation and search use Google APIs.) Thanks to the Nemoclaw layer, API keys and similar secrets are kept hidden from the AI, which makes the system reassuringly secure. I didn’t write a single line of code myself. The AI wrote the code, tested it, and then organized it so it can be reused. It also writes the reusable prompts and skills. It’s not quite what you’d call full orchestration yet—the main agent issues instructions to sub-agents. Managing the context window limitations, along with prompt optimization and compacting, turned out to be key. It even wrote articles that I could learn from myself, so I’d call that a pass.

I’m been coding for more than 30 years and I forbade myself from coding for a week so locally running LLM agent does. I found that it takes great deal of computing knowledge to make agents produce reliable works efficiently.

A long stdout burns tokens. Instruct AI to dump into a file, then use head/tail/grep.

I need to improve the agent observability...just relying on nvtop to make sure AI agent hasn't stopped isn't a sophisticated approach 🙄

Harsh comments are nothing when you post about something you are so convinced.