HeyJaredLer Tech Notes
Instructor of AI Agent classes | Programmer since 2010 | Blogging since 2007 | Helping working adults and students build practical AI tools | Exploring ethical AI writing, automation, and education.
- Indexed issues, last 90 days
- 14
- Latest publication
- Sep 18, 2026
- Audience
- Checking…
- Earliest in this view
- Aug 7, 2026
Latest issues
Can Singaporeans Be the Real Market for AI Classes in Johor Bahru? (opens the original)
Read excerpt
Recently I started asking a different question.Instead of:“Why is it so hard to fill physical AI classes in Johor Bahru?”Maybe the better question is:“Can JB serve Singapore learners instead?”There are some interesting data points.In 2025, Johor recorded 24.7 million international visitor entries, including 19.38 million visitors from Singapore. Singapore remains Johor’s largest visitor market. And cross-border travel may become much easier soon.The Johor Bahru–Singapore RTS Link is
Stop Comparing Claude, Codex, and Gemini. Build a Better Harness Instead. (opens the original)
Read excerpt
1. Treat AI models as replaceable commoditiesMost organizations spend too much time comparing:Claude vs CodexGPT vs GeminiOpenAI vs AnthropicThe author’s argument:Model advantages are temporary.Today’s best model may not be next month’s best model.Performance gaps change rapidly.Strategic takeaway:Build workflows that survive model changes.A good system should allow you to swap models without rebuilding everything.2. Reliabi
How I Create a 10-Second CTA Video with Claude Code and Codex (opens the original)
Read excerpt
1. Start With the Marketing Problem, Not the AI ToolI was preparing to conduct an AI for Dhamma Outreach workshop at d’CRADLE on 3 October 2026.The initial response was slower than expected.One of the organisers gave me useful feedback: instead of relying only on posters and links, why not create a very short Reel that immediately shows how the workshop can make video editing easier and reduce repetitive work?That changed the question.Instead of asking:“Claude, create a promotional video for my
The Real AI Race Is No Longer About Models. It's About Services. (opens the original)
Read excerpt
For the past few years, most AI discussions focused on models.Who has the smartest model? Who has the most parameters? Who scored higher on the latest benchmark?That conversation is becoming less useful.After watching a recent discussion featuring Ji Gang, GuiZang, and Ju Zi, one idea stood out: the winners of the next AI wave may not be the companies building the largest models. They may be the companies building the best services, infrastructure, and ecosystems around those models.The model is
Why Is Colibri's DeepSeek V4 Flash So Slow on Consumer Hardware? A Real Benchmark From Production (opens the original)
Read excerpt
Most discussions about large MoE (Mixture-of-Experts) models focus on GPUs. This one does not.I tested DeepSeek V4 Flash running under Colibri on a consumer server with a single NVMe SSD and no GPU acceleration. The result was surprisingly consistent:Time to first token: ~38 secondsGeneration speed after the first token: ~1 token every 6 secondsTotal response time: ~70 secondsMeasured across five real requestsStreaming enabled and verified<
Publishing over time
Last 90 days. Choose a month to open its work.
Recurring subjects
Named in the text we hold. One piece can cover several.
Audience
No verified audience measurement yet.
About this data
Counts cover the work we have indexed. Tone needs enough text and a confident classification. Excerpts and episode notes are not full articles or transcripts.
Identity or attribution wrong? Suggest a correction.