I created a daily investment digest customized for my portfolio showing current and 30-day market trends, key news, upcoming IPOs/Earnings and correlating these to provide short/long-term investment guidance. One thing that required fixing was the indices returning incorrect %change. Yahoo Finance's API doesn't return a clean "today's change" field — regularMarketChange comes back as null, so the original code fell back to chartPreviousClose to calculate the change. The problem: chartPreviousClose is the closing price from before the 30-day chart window (i.e., ~31 days ago), not yesterday's close. So the math was wrong — it was comparing today's price to a month-old price. The fix was to ignore those fields entirely and use the raw candle array instead: - closes[closes.length - 1] = today's live intraday price (Yahoo always appends today as the last candle) - closes[closes.length - 2] = yesterday's actual close Subtracting those two gives the correct day change.

This was a great intro to Claude Code, and has already inspired many use cases! Understanding what claude.md does was the key to unlocking this new coding interface. I quickly ran out of API credits going through the iterations. A takeaway would be to explicitly tell the program to ask whether to reuse data (eg. Perplexity results, nano banana images) each time. I also ended up handing over image generation to Python's Matplotlib instead to keep things moving without investing more kie.ai credits during the learning process. Pardon my poor use of colors (Agenta -> Magenta), just wanted to get something up and running quick.

As a building block towards an automated YouTube channel, I've created the following skill and a partial workflow: Name: /history-narrative Description: Based on user's desired topic (eg. Rise and Fall of the Byzantine Empire) and duration, scour the web with perplexity and generate a narrative. Writing style references a guideline to keep audiences engaged, including planting tension pivots and curiosity gaps every ~12s, and using the personality of my channel. Adjustments: - The first iteration had misleading wording with the intention of creating a hook. The skill was updated to ensure information is factual and not prone to misinterpretation. - Style guide was extracted into a global narrative-style.md reference file so that it can be re-used for skills dedicated to other topics (eg. travel, politics, stock market, etc.) Output 1. 5-min narrative on the Byzantine Empire following guidelines (attached opening here only) 2. The narrative is then used to plan graphics, get approval from user, before generating images via nano banana 3. Images are fed to an open-source /whiteboard-animation skill to create whiteboard animation Next steps 1. Simplify graphics and adjust timing of whiteboard-animation to keep rendering easy on the eyes 2. Created /map-generation skill to represent armies moving/engaging -> to be optimized, using HTML/CSS/JS for now 3. Looking into Nate's resources on hyperframes to automate video creation, remember preferences, and evolve Can't wait to see the first video come to fruition!

For this exercise, I scraped the US government public bids website (sam.gov) for listings matching our products' NAICS code over the last 5 days. Claude Code really deserves full credits on this one! Interestingly, it determined from the outset that using sam.gov 's API was a better option than Fire Crawl due to the many authentication layers on sam.gov. Upon using the API, it then discovered that the keyword filtering parameter was not working, and turned to Fire Crawl to cross-reference the website's search fields to determine other possibilities. Turns out that the API can only return a date range of data, with no filtering capabilities. We decided it was best to fetch the data, then perform keyword filtering offline. This scraping performed periodically can be packaged into a weekly leads report for our sales team to follow up on.

Let's get to know each other! Comment below sharing where you are in the world, a career goal you have, and something you like to do for fun. 😊