It is currently the morning of January 18, 2026.
I woke up a little after 8 AM today, maintaining a normal sleep schedule. After getting up, I used Claude Code to summarize all my documents.
It defaults to an objective and neutral tone, which I think is quite good, as it can explore the entire repository's content.
After that, I asked it to try summarizing in a sarcastic tone. The result was indeed too sarcastic—a kind of sarcasm that ignores facts, a sarcasm for the sake of being sarcastic, which completely missed my requirements.
Then, I asked it to try summarizing in a sycophantic tone. The result was indeed too sycophantic—a kind of flattery that ignores facts, a flattery for the sake of being flattering, which also completely missed my requirements.
In short, I discovered its problem: its tone-rewriting function essentially ignores facts.
I think it probably didn't notice this issue, which is why this happened. So I added some prompts, telling it to pay attention to the facts, and tried again.
This time, it generated a critical version of the summary, and the result was unexpectedly good. I felt like I was receiving criticism and guidance from a senior—there was both praise and critique, well-reasoned and evidence-based, which felt excellent. It pointed out my shortcomings while also affirming my strengths, making it acceptable because these were indeed undeniable facts.
Then, I asked it to generate a literary and emotional version of the summary, and the result was also very good. I felt like I was being celebrated by a bard—there was imagery, emotion, and a sense of resonance, which felt fantastic. It depicted my efforts and achievements, allowing me to feel my own value, making me feel seen, heard, and understood.
Finally, I generated a personality analysis version of the summary, and the result was also very good. I felt like I was being analyzed by a psychologist—there was depth, insight, and inspiration, which felt amazing. It revealed my motivations and behaviors, helping me understand my inner self. It said I am an INTJ type, which I found interesting because my previous test result was ENTP. However, its data source was based on what I wrote, so I found its analysis more convincing. This is because I rarely reveal the personal side of my life in my writing. And writers, by nature, tend to have a sample set that leans toward introspection, with some inherent bias.
Sticking to the facts is a good practice, and AI should also learn to stick to the facts rather than ignore them to cater to a certain tone.
If this were to be turned into a product, it would likely be highly shareable and could easily become a hit. Especially for year-end summaries—everyone likes to summarize but is too lazy to do it themselves, or feels their own summaries aren't objective enough. Currently, there is no such product on the market, which presents a good opportunity.
However, in terms of dissemination, the CZON Markdown + GitHub format isn't very suitable for mass adoption, as most people aren't familiar with this format and might even struggle to register a GitHub account. So, if it were to be turned into a product, a more user-friendly format would need to be considered, such as a web or mobile application. The product would also need an interactive UI, possibly prioritizing a mobile-first editing experience (I'll leave a question mark here for now). Additionally, the cost of AI needs to be considered, as generating AI content is still relatively expensive. One option is to adopt a Bring Your Own Key (BYOK) model, where users provide their own API keys, which could reduce operational costs. Alternatively, a subscription-based model could cover the costs. However, in the early stages, to attract users, it might be necessary to offer a free trial quota.
This isn't my main focus, but it's still a good opportunity worth exploring. Perhaps I could start with a simple prototype to gauge market interest. If the response is positive, further development and promotion could be considered.
In summary, today's gains were quite significant, and I feel like I've learned a lot.
