Today was lighter and more structured.

After reviewing a demo, I received clear feedback on how to improve the MCP experience and storytelling around our AI workflows. I documented key LLM thoughts to prep for tomorrow’s demo and help future planning. In the afternoon, I joined a discussion on improving feature service performance.

Back home early, I enjoyed dinner and a rare walk with Xiaoyan. She’s close to sealing her first sales deal—something she sourced and pursued herself. Seeing her so thrilled made me feel genuinely proud and happy.

Today was a mix of stress and support.

I accompanied Xiaoyan for her full-anesthesia endoscopy—she was nervous and starving after fasting since the night before. Luckily, results came back mostly normal.

At work, I had back-to-back meetings prepping for Friday’s demo and team’s AI sync meeting. I felt the pressure, especially when struggling to explain some workflow. But progress was made—we finalized the UI and backend in the afternoon.

To decompress, I played badminton and spotted a gorgeous streak of fire clouds on the way home. Exhausted, I crashed early.

I woke early for a 6 a.m. meeting after a rough night—my electric toothbrush kept turning on by itself and disrupted my sleep. The meeting felt uneven: the rushed presentation lacked substance, while the live demos sparked real discussion. This highlighted an ongoing issue—pressure to deliver quickly, without proper planning.

At work, I spent most of the day debugging MCP backend functions. By evening, I was completely drained. Xiaoyan came home early to prep for a medical checkup.

Today was a high-pressure, emotionally heavy day. I struggled to make progress on the AI Assistant workflow, despite looming deadlines. A sudden meeting assigned me leadership, adding to my stress. On top of work, we received a critical health notice about my uncle, which deeply shook me. Thankfully, he’s stable for now. The mix of professional pressure and personal concern made it a tough day to navigate.

We started the day groggy and late after a long, tiring Saturday. Xiaoyan had a hospital visit for a digestive checkup, and a minor misunderstanding between us was quickly resolved with an apology. In the afternoon, we visited Jialing’s new house for a cozy lunch and dinner.

Today was emotionally heavy. We visited Xiaoyan’s former colleague in hospital. In the afternoon, Xiaoyan handled appointments while I prepped for soccer. We also discussed getting critical illness insurance for Hamer—a decision influenced by today’s visit. The evening’s match was hot and tense, ending early after a foul nearly led to a fight.

Today felt emotionally packed—from team pressure and unclear AI tasks at work to hearing heartbreaking news from a friend. I tried outlining Earth AI Assistant’s tech structure but hit some roadblocks. Lunch with a colleague brought reflection and warmth. It was also Hamer’s birthday—seeing family gather for her made me smile. The evening turned heavy when Xiaoyan shared her fcolleague’s cancer battle.

Today started with a home network upgrade—finally moved the router to the living room for a smoother TV and HomePod experience. My windowsill workspace is also shaping up nicely.

At work, I focused on MCP improvements and later joined a two-hour AI discussion on AgenticFlow. A colleague’s farewell lunch added a warm midday pause.

In the evening, I played badminton at a slightly dated sports center—decent facilities, low-key vibes, and just enough sweat to call it a day.

Today was a whirlwind—excitement over an R&D spotlight turned into frustration as Earth’s session was merged into a broader Beijing showcase. I was then asked to re-present Earth AI Assistant in a long internal meeting. With little prep time, I focused on Semantic Kernel and demoed key workflows. Thankfully, the response was positive.
By day’s end, I was completely drained—no lunch, constant pressure, and my first-ever three-hour internal meeting.

After a night of rain and a day full of meetings, I found myself somewhere between progress and pause. I explored Zhipu’s GLM-4.5, wrapped up lingering MCP tasks, and juggled overlapping pressures from both the Earth and AI teams—all aiming for September results. A dream about a collapsed bathroom somehow mirrored that tension. Tonight, I chose not to push further. Sometimes, it’s okay to pause and leave a little room to breathe.

On yet another rainy day in Beijing, I dove into ideas around making AI Build support custom LLM API keys—but Semantic Kernel remains limited to OpenAI. I also explored ChatGPT Agents (visually slick, but less interactive than Claude Code).

A sudden insight hit while unwinding with Xiaoyan: our Agentic GIS might align beautifully with Claude Code’s Subagents model. Fired up, I explored it late into the night.

Meanwhile, my brother called—Hamer’s swimming over 20 meters now. She’s growing fast.

We spent the entire day decluttering—mainly tackling two balconies and tossing out loads of unused items. Rearranging furniture and switching to fresh IKEA curtains completely transformed the space, making it feel lighter and calmer.

I sold 55 books online for ¥168 after shipping costs and gave the rest to our building’s cleaner. We cleared nearly 100 books in total. It was exhausting but deeply satisfying—more joyful than a day out.

We spent the entire day decluttering—mainly tackling two balconies and tossing out loads of unused items. Rearranging furniture and switching to fresh IKEA curtains completely transformed the space, making it feel lighter and calmer.

I sold 55 books online for ¥168 after shipping costs and gave the rest to our building’s cleaner. We cleared nearly 100 books in total. It was exhausting but deeply satisfying—more joyful than a day out.

I started the day chasing ideas to improve my handwriting app—especially for pinyin and stroke parsing. Found some solid resources, but work came calling.

Spent the day refining the Identify logic with better parallel handling. After dinner, I updated the app using those references—progress made, but it’s still rough (and Claude API credit is running low). I was really happy to finally discover the clever loading word indicator design of Claude Code.

Render Hanzi writer character data, stroke animations or stroke quizzes. Hanzi Writer

• chanind/hanzi-writer: Chinese character stroke order animations and practice quizzes

skishore/makemeahanzi: Free, open-source Chinese character data

• Learn Chinese one stroke at a time

HSK levels related

• Word Lists - HSK东西

I was attracted to the word loading indicator when Claude Code is working on a response. For example, jiving, whirring, channelling, conjuring, pontificating, concocting, discombobulating…

After searching from the web, I found this topic Claude Code’s loading messages which resolved my curiousity.

…found out that Claude Code will actually call the API and ask Claude Haiku to generate this single word based on your input. It does this as you type so it’s ready to go when you submit (and it will make many calls for each prompt).

For those interested, here’s the full system message for the request. It will send you input in a separate user message.

Analyze this message and come up with a single positive, cheerful and delightful verb in gerund form that's related to the message. Only include the word with no other text or punctuation. The word should have the first letter capitalized. Add some whimsy and surprise to entertain the user. Ensure the word is highly relevant to the user's message. Synonyms are welcome, including obscure words. Be careful to avoid words that might look alarming or concerning to the software engineer seeing it as a status notification, such as Connecting, Disconnecting, Retrying, Lagging, Freezing, etc. NEVER use a destructive word, such as Terminating, Killing, Deleting, Destroying, Stopping, Exiting, or similar. NEVER use a word that may be derogatory, offensive, or inappropriate in a non-coding context, such as Penetrating.

This is what the official docs have to say about it:

Haiku generation: Small creative messages that appear while you type (approximately 1 cent per day).

Xiaoyan was drained from a full day of pricing negotiations, but I saw it as a sign that deeper collaboration might be brewing. We shared dinner and laughs tonight. Hamer called saying she missed me. She’s now learning to swim at the swimming pool.

At work, I wrapped up bug fixes and cross-testing ahead of schedule and resumed using KanbanFlow.

We caught the midnight screening of The Stage by Chen Peisi—a film adaptation of the stage play I had long wanted to see. It was funny at first, then slowly grew emotionally gripping, reminding me of Farewell My Concubine. One standout was Feng Xiaotong’s portrayal of Yu Ji—subtle, moving, unforgettable.

Earlier in the day, we met friends for lunch and checked out their new NIO EV. After a ride demo and some good food, we ended the day with great company and a powerful film.

Today was a mix of focus and small joys. I tackled threading issues in Identify and knocked out several bugs—always a satisfying rhythm. Scored discounted shoes at MUJI and enjoyed a cozy tofu soup dinner.

Hamer called a few times just to chat, showing both her need for connection and her growing thoughtfulness.

Xiaoyan left for Qingdao and reconnected with an old friend. I ended the night doing nothing in particular, just watching Goodboy TV series past midnight. Sometimes, that’s enough.

I spent the day chasing down a sluggish Identify popup issue—and finally cracked it with a clever use of ConfigureAwait(false). Along the way, I tested using AI tools (Cline + Kimi K2) to spin up a WPF app, which turned out smooth thanks to my familiarity with the platform. Compared to building an iOS app yesterday, this was far more efficient.

The day ended with some badminton, dinner with family, and a relaxing video scroll.

Before bed, I finally used my new SSD to back up files from both my and Xiaoyan’s computers. While cleaning things up, I realized how little I’ve organized over the years. Just browsing my Downloads folder felt like opening a digital time capsule—messy, but filled with memories and traces of everything I’ve done.

After a restless night and an early team meeting, I spent the day tracking a tricky performance issue and brainstorming a long-desired personal project. In the evening, I tested building an app using AI tools—Cline+Kimi K2 failed to deliver, but Claude Code exceeded expectations with a solid, well-structured output. I also enjoyed its confidence and quirky vocabulary like discombobulating. A $5 spend felt worthwhile.

TIL: iOS App not using full height of the screen

Problem:
iOS app is not taking the whole height of the screen.

Reason:

“where I removed … the launch screen”
All iOS apps MUST have a launch screen storyboard, even if the rest of the app is done without storyboards.

Solution:

1. Go to the target’s build settings
2. Search for Launch Screen (Generation)
3. Turn the value to Yes

I started the day feeling oddly anxious and had a disturbing dream, but things turned around by evening. After a productive afternoon recording an MCP demo and helping Hamer learn how to send a video on her own, the real highlight was paddling on the Liangma River with Xiaoyan. Drifting on the calm, dark water beneath city lights and a soft breeze was pure peace—two hours of simple joy that erased all the day’s stress.