Today, I felt utterly lost—eager to accomplish everything yet ending up doing nothing meaningful.

During the work, I tackled PM questions in the moring and attended my first Dev Lead meeting. I struggled to balance team updates with my own ideas about an internal AI weekly newsletter.

At home, I cooked dinner with Xiaoyan, sharing a brief moment of culinary success. However, the day ended on a sour note when Hamer, upset over a broken promise regarding her new underwear, erupted in tears and a tantrum, triggering a conflict between her and Xiaoyan. Despite no further escalation, I was left feeling burdened by unresolved resentment.

TIL: Top AI News of Week 15

💻

AI 2027

This website is a detailed scenario analysis predicting the potential impact of superhuman AI. Beside the content, this website provides a very good UX, the whole page is interactive. You can find the right side graph dynamic updates while scrolling.

AI 2027 is a comprehensive and detailed scenario forecast of the future of AI. It starts in 2025 and projects the rise of AI agents by 2026, the complete automation of coding in early 2027, and the intelligence explosion in late 2027. It has two branches, one ending in AI takeover and another ending in utopia (sort of).

Making AI Actually Work on Your Team

The article emphasizes the importance of AI in modern software engineering teams and provides practical advice for engineering leaders on how to integrate AI tools effectively. It highlights the need for leaders to guide their teams through this transition and create a supportive environment for experimentation.

Seeing is believing First things first: try it yourself. … So get your hands dirty… Your personal experience matters here—without it, it’s all just theories in your head; you’ll be out of touch. … I gave him/her the honest truth: “It’s impossible for me to promise job security for anyone. But think about this logically: we hired you before AI was a thing, and now that you’re significantly more productive with these tools, why would we let you go?”

Companies don’t (usually) fire their most productive people; they invest in them. The real threat isn’t AI—it’s sticking to outdated ways while the industry evolves around you. … The shift to AI-assisted engineering isn’t coming—it’s here.

The Dynamic Between Domain Experts & Developers Has Shifted

The article discusses how the relationship between domain experts and developers has changed in the AI era, emphasizing the increasing importance of domain expertise in building successful AI products. It’s interesting to see the sentence and picture of “I’ve got a great idea for an app…”.

… But today, I’m seeing teams of domain experts wading into the field, hiring a programmer or two to handle the implementation, while the experts themselves provide the prompts, data labeling, and evaluations.

For these companies, the coding is commodified but the domain expertise is the differentiator.

229 things we announced at Google Cloud Next ‘25 – a recap

Google Cloud Next 2025, held in Las Vegas, showcased over 200 announcements emphasizing advancements in AI, infrastructure, and real-world applications.

Here’s a concise summary of AI & Machine Learning:

• Gemini 2.5 Pro: Introduced as Google’s most advanced multimodal AI model, now in public preview across Vertex AI, AI Studio, and the Gemini app.
• Agent Ecosystem: Launch of tools like Agentspace, Agent Development Kit, and Agent2Agent protocol to facilitate the creation and management of multi-agent AI systems.
• Gemini Code Assist: An AI-powered coding assistant integrated into development workflows, offering features like a Kanban board for task management and insight into AI decision-making processes.

Google Is Winning on Every AI Front

The article argues that Google and DeepMind have taken a significant lead in the AI landscape, outpacing competitors like OpenAI, Anthropic, and Meta. The author, who previously had high hopes for OpenAI, now believes that Google’s strategic and technical advancements have positioned it as the dominant force in AI.

(PSA: Many people are interested in this post, so I removed the paywall)

With strong winds forecasted, we spent the weekend cozied up at home—it felt like a mini vacation.

On Saturday, I slept in late after a night of binge-watching TV. Hamer kept whispering in my ear until I finally got up. “This is the first time I’ve seen Dad get up this late,” she said. I lost a bet with Xiaoyan over the pronunciation of “divorce” and had to make lunch—two veggie dishes and a fried egg that turned out surprisingly well. Later, Hamer and I continued our series binge, and even Xiaoyan joined in after dinner, crying during several scenes. That night, I slept soundly in Hamer’s bed.

Sunday was equally low-key. After playing in the morning, we went grocery shopping for braised green bean noodles. Hamer also wanted a new watercolor set because hers lacked a skin-tone color. There were a few tears during the discussion with Xiaoyan, but she got it in the end.

In the evening, Xiaoyan cooked radish and cuttlefish soup—it was delicious. Hamer helped with folding laundry, hanging clothes, and washing fruit. She’s really becoming Mom’s little helper.

Meanwhile, I got deeper into pour-over coffee and finally dialed in the grind size. It’s starting to click.

Last night’s badminton game left me with a twisted ankle, and by morning, it was still sore and uncomfortable to walk on. After dropping Hamer off, I grabbed McDonald’s for breakfast—just had a craving.

Later, I visited the dentist to follow up on a previous check-up. X-rays revealed an old filling that may need a root canal. For now, I had a cleaning and got a reminder: no horizontal brushing, use an electric toothbrush, floss well, and pay extra attention around my implant.

In the afternoon, I worked on setting up an old AI project environment. With help from a colleague, I finally got it running and plan to document the process and improve it next week.

Before heading home, Hamer called and joked that I might get blown away if I didn’t hurry. After dinner, we watched When Life Gives You Tangerines and only realized IU was the lead by episode two.

The long-anticipated windstorm still hadn’t arrived by bedtime—maybe tomorrow.

Want to share or archive your GPS tracks from ArcGIS Earth? Exporting your NMEA layer as a KML file is quick and easy.

📁 Simple Export:

After importing and visualizing your NMEA log, right-click the layer and save. You’ll get:

• Points representing individual GPS fixes.
• Polylines connecting your track path.

It’s the perfect way to share your routes or archive valuable tracking data.

🚀 Coming Soon: Enhanced KML Tracks:

Currently, ArcGIS Earth exports points and lines to KML. Stay tuned for full KML Track support, including timestamped playback and richer temporal visualization!

✨ ArcGIS Earth empowers you to not only review your GPS logs but also transform them into easily shareable and presentation-ready KML files.

2025-04-10 TIL: AI Model Performance Analysis

💻

AI Model Performance Analysis

This analysis highlights the tradeoffs between model size, speed, and tool-calling capabilities across different AI models. Llama3.1:8b appears to offer the best balance of features, while llama3.2:3b provides a good compromise between size and functionality.

Hamer’s been waking up early lately, getting ready calmly and playing on her own before school. After dropping her off today, I grabbed breakfast, ate wontons while waiting for corn, and brought extra food home.

In the morning, I subscribed to CapWords—originally just out of curiosity, but the heartfelt story behind it deeply resonated with me.

I spent the day testing local language models with the Earth MCP Server. A larger model was too slow, and a smaller one underperformed, but I eventually got two working and pulled together some solid takeaways. Before wrapping up work, I submitted a documentation PR to the oterm project—my first open-source contribution—and was thrilled to see it merged right away. I hope it’s just the beginning.

Around noon, I was notified of a seatbelt violation from last Sunday. It surprised me—I’m usually careful—but the photo said otherwise.

All the talk today was about Beijing’s strong wind forecast. Looks like we’re staying indoors this weekend.

Think your GPS tracks are just lines? Think again! ArcGIS Earth empowers you to dive deep into the metadata, attributes, and statistics embedded within your imported NMEA layers, revealing the full story.

📄 Explore Layer Details:

Right-click the NMEA layer and select “Properties” to access:

• Metadata: spatial reference, extent, data source, and more.
• Summary Statistics: a quick snapshot of your dataset’s structure.

🖱️ Instant Feature Insights:

• Use the “Identify” tool to click any point or segment and instantly view its attribute information in a pop-up. Perfect for on-the-fly checks during playback or exploration.

📊 Dive into the Data Table:

Open the “Attribute Table” to:

• See all your GPS records in one comprehensive view.
• Sort, filter, and analyze your data effectively.
• Select specific records and zoom directly to them on the map.

✨ Go beyond the lines! ArcGIS Earth provides the tools to unlock the rich contextual information within your NMEA data, transforming simple tracks into powerful insights.

I started the day by continuing to listen to the ongoing company meeting—by this point, I had followed it from start to finish, which made me feel somewhat involved in the whole session.

Throughout the day, I explored integrating Ollama with MCP to run local language models. I initially tried Open WebUI’s mcpo to expose MCP as an OpenAPI interface, but I didn’t know how to use the tools it offered. Later, I found oterm, an Ollama client that supports MCP tools. When I couldn’t select tools, I realized that not all models support the tools feature.

When we got home, Hamer surprised me by pretending to be a waitress—seriously guiding me through removing my shoes, washing up, and serving dinner. Her commitment to the role was both hilarious and sweet. After dinner, we went shopping. I was impressed by how disciplined Hamer has become—carefully choosing only a few things. She’s really growing into a thoughtful, self-controlled little person.

Imported an NMEA log into ArcGIS Earth? Awesome! Now, transform it into a powerful visual story by customizing its appearance for clearer analysis and compelling presentations.

🗺️ Control Your View:

• Right-click the NMEA layer in the Table of Contents and select “Display.”
• Choose to show Points, Polyline, or Both to focus on key aspects of your track.

🎨 Style for Insight:

• Treat your NMEA layer like any feature layer – unleash the power of symbology!
• Apply color ramps to highlight speed, accuracy, or elevation variations.
• Adjust size and style for optimal visibility in immersive 3D scenes.

🏷️ Label for Context:

• Bring your data to life by labeling attributes like timestamp, speed, or elevation.
• Enhance analysis and presentations with crucial contextual information.
• Add dynamic real-time context to your visual playback.

✨ Unleash the full potential of your NMEA data! By taking control of display and styling, you transform raw data into insightful and visually captivating narratives.

I went to bed late last night after an online company meeting, and when I woke up, it was still going. I tuned back in and was struck by how open and confident everyone was, regardless of role.

The rest of the morning was filled with more meetings, but our team meetings felt flat—minimal interaction and little new information. It’s something I’ve been reflecting on.

I spent the afternoon preparing NMEA data visuals for a presentation and began exploring LM Studio and Ollama’s Python SDK to see how local models might support MCP tasks.

On the way home, I got my badminton racket restrung, filled a growler with beer, and grabbed barbecue for dinner.

Xiaoyan and I also considered renting a car for a weekend trip now that I’m confident driving, but the windy forecast led us to postpone until next week.

Did you know that ArcGIS Earth not only replays NMEA GPS tracks—but also lets you explore detailed positioning and satellite data at every moment?

📍 Go Beyond the Map with Location Information When working with an imported NMEA log, click the Location Information button to open two insightful tabs:

🧭 Location Details

Access key metrics about the device’s current position:

• Latitude, Longitude, Altitude
• Accuracy level and Fix type
• HDOP (Horizontal Dilution of Precision)
• VDOP (Vertical Dilution of Precision)
• PDOP (Positional Dilution of Precision)

These details help you assess GPS signal quality and positioning precision throughout the track.

🛰 Satellite Information

Visualize the satellite environment in real time:

• View the relative position of each satellite in the sky
• Signal strength is displayed with dynamic bars—longer and darker = stronger signal
• A compass-style map shows satellites orbiting around the device at the center

✨ This view updates continuously, giving you a real-time window into satellite coverage and positioning accuracy.

Did you know ArcGIS Earth supports NMEA log file import, enabling you to visualize and analyze GPS tracks with ease?

📡 What is NMEA 0183?

NMEA 0183 is a widely adopted standard developed by the National Marine Electronics Association. It defines how GPS devices and marine instruments—such as sonar, anemometers, autopilots, and more—communicate using simple ASCII “sentences” over serial connections.

It’s the most common GPS data format, supported by virtually every GPS manufacturer.

🗺️ ArcGIS Earth Makes It Visual

Simply import a NMEA-formatted log file into ArcGIS Earth to bring your movements to life on a map.

🔁 Playback & Insight at Your Fingertips

Once imported, select the NMEA layer to open the GNSS Toolbar and Dashboard:

• ▶️ Play or pause the GPS track
• 🐇🐢 Adjust playback speed in real time
• 🧭 Choose your view mode: Track-Up, Keep at Center, or Keep within Window
• 📊 View live stats: current speed, total distance, elevation gain, total time, and average speed
• 📈 Monitor progress with an interactive elevation profile

🎯 Real-World Application

Whether you’re reviewing a field mission, validating GPS data, or training with real-world tracks, this feature transforms raw GPS logs into powerful, interactive stories.

This morning, Hamer woke up teary-eyed, missing Liuliu deeply. She cried as she got ready but gradually calmed down.

After the dev meeting, I reflected on how to make these sessions more impactful.

In the afternoon, I settled on NMEA playback as the focus for this week’s Earth 230 Tips, which gave me clarity and momentum. I also completed paperwork for my work-residence permit renewal—much easier now that the company is handling it directly.

We decided to skip this week’s family badminton session. Xiaoyan felt I’ve been too critical during play, and Hamer said she was a bit tired. After dinner, we took a relaxing evening stroll and picked up some fresh fruit.

Today, I drove back to Beijing, covering 250 kilometers in 7 hours, bringing my total driving distance over the last two days to 400 kilometers—the longest I’ve ever driven. It was a challenging but rewarding experience, and by the end of the day, I truly felt like a “real driver.”

We left around 11 a.m. and got home by 6 p.m. In the evening, we celebrated my mother-in-law’s 62nd birthday with a family dinner. Happy Birthday, Mom!

Today, I drove on the highway for the first time—about 150 kilometers. I started out nervous but gradually found my rhythm, and by the end, I felt proud and accomplished.

In the morning, we visited Huanghua Port and Golden Beach, where the wide river and sandy shores created a relaxing coastal feel. Hamer and Liuliu played in the sand while we strolled and chatted under the sun.

In the afternoon, I drove Xiaoyan and Zhu Jian to Cangzhou West Station, tackling both local roads and the highway. The 100-kilometer stretch was intense at first, but I settled in and gained real confidence behind the wheel. Returning solo to Liuliu’s house, I felt calm and capable.

That evening, the warm, everyday scene—Liu Dan in the kitchen, the kids playing—filled me with comfort and gratitude. Their family’s kindness and support made me feel truly at home.

We got up at 5 a.m. to head to Huanghua, where Liuliu’s dad drove us to visit her at her grandma’s place. Hamer and Liuliu were clearly thrilled to see each other, though they tried to play it cool. I was unsure about driving back to Beijing alone and hoped Liuliu’s dad could take the car while we returned by train.

Lunch was a feast—seafood, lamb spine hotpot, and shrimp paste flatbread—that left me full well into the evening. We strolled through Xinyulou in the afternoon, and later, the four of us played mahjong until past 11. Afterward, Zhu Jian took me visit a local bathhouse for a soak and steam, spending the night there.

Tonight, we attended our first performance at the China National Opera House—a symphony and ballet set to The Yellow River. The music, staging, and dancing were breathtaking, and we left wanting to see it all over again.

Since our e-bike couldn’t fit four, I planned to run home—but Hamer wanted to join. We waited backstage, caught glimpses of the dancers leaving, then walked home together, snacking on frozen grapes from 7-Eleven. We held hands, laughed, and skipped until she got tired. At Chuangshan Liquor Bar, we stopped for skewers and beer to rest.

After dropping Hamer off at school this morning, I grabbed a big breakfast to share at home—more than we could finish.

Most of the day was spent diving into the MCP C# SDK. While reviewing the official samples, I spotted a bug I initially planned to fix, only to find someone had already submitted a PR—so I added a supporting comment instead. I also spent time exploring how to integrate MCP with the Earth project but realized it’s not easy to integrate directly.

In the evening, After shopping, we visited a print shop to print photos for Hamer.

Hello, April.

This morning, I woke up early and joined Hamer for a morning run. We ran and walked around 3 kilometers together—it was a joyful experience. On our walk back, she said the day felt like a weekend without school.

At work, I learned I’ll be taking over the lead of a project. It didn’t hit me until I was added to the group, making me realize this new responsibility.

In the evening, I grabbed some draft beer and skewers on a whim, and Xiaoyan couldn’t resist nibbling despite being full.

There was a moment of misunderstanding with Hamer’s homework, where I was adamant I was right, but later discovered she was correct. I realized I need to listen to her more and not always assume my perspective is the right one.

We also got our registration confirmed for the parent-child run at a local Half Marathon. Looking forward to it!