Nick Wverified
@nickw_ai
Learning AI Engineering
Score
19 / 375
Day
59 / 75
Streak
1day
All 75 days
2 perfect5 activelongest streak 3
Posts
Hopped on a Call with Sophie Liu
I just spent $100 for 45 minutes with a Chinese life coach, and here's what it taught me: People are the biggest unlock. At the end of the day all life is, is interacting with other people, the job you want, the business you want to create, the community, connection, relationship you're looking for. It's all about talking to people, helping them, convincing them, teaching them, learning from them - cultivating those relationships and from that gaining some degree of truth or perspective or understanding about the world. So for me, as someone who wants move to a different country, explore entrepreneurship, all these different things. Hoping on this call gave me the advice I needed to hear to make the best decisions possible and also connect me to the people that I could portentially build this. So if there's anythgin to take away from this, it's to talk to more people, make more friends, "network" and "coffee chat" more.
Wed 26 Aug · score 0
Claude API + Scoping out Sourcing Agent Build
I finally got to the tool section of the Claude API course, and scoped out what it'd look like to building a sourcing agent. Sourcing is essentially a process of researching what product you want to produce, finding factories that can produce said product for you, reaching out to them to do business which can involve getting samples from multiple, comparing the quality and prices/quotes you get from them, and then finalizing/negotiating terms. All in all, it's very involved process that could take several weeks to months to bring a product from idea to finished product. The idea behind an AI sourcing agent is that it can handle many of these steps autonomously, speeding up the process and freeing up the time of the person sourcing.
Tue 25 Aug · score 0
Started the Build w Claude API Course Today
learned about evaluating prompts and what adjusting the temperature of a model means (closer to 0 is more deterministic, while closer to 1 make the probability distribution of the next set of tokens more spread out, increasing the likelihood of "out there" outputs)
Sat 22 Aug · score 0
Day 59 of 75