Things read - Week 30
Kevin Kelly's Profile
My obsession with Kevin Kelly isn't new here. He is someone I wanna become after I grow up. So naturally, when this profile came up diligently read it. And apart from his usual principles of following your interests obsessively and compulsively without caring about the money, I was floored when he says Greatness is overrated. It's a form of extremism and it comes with the extreme vices that he is not interested in. The author mentions that when he (the author) forced himself to come up with a professional goal, all he could muster was "have a good day, most days." This is the essence of a life well lived. It's not entirely related but in the book 'Remains of the Day', the protagonist reflects on his life and says something along the lines of how evenings are the best part of the day because once you did justice for the days (i.e., using your time well), we can enjoy the evening guilt free.
Another interesting thought KK mentions that he doesn't pursue a destination but rather a direction. It sounds fascinating because with this we can allow ourselves to not be side-blinded by other interesting things that might be useful for us to pickup. Probably this is the reminder that I should pick up: Why Greatness Cannot be Planned. Flounder Mode
AI skepticism
With all the AI wave (and honestly having sooo much fun riding it), this story was a bit of fact-check of the hype. Although, I did read with some skepticism it was good to be reminded that essentially all models are a only a bit more than 'stochastic parrots.' An interesting phrase that she coined in 2021 which means a system “for haphazardly stitching together sequences of linguistic forms it has observed in its vast training data, according to probabilistic information about how they combine, but without any reference to meaning." She argues that all the data that these LLMs are trained on are flawed from the inception because the data sets used are already biased. This is more of a critique of the society than the models but still useful to be reminded. Also, bonkers that we are expecting these models to act as therapist and friends and equally bonkers that we believe it when it says it understands our problems. AI sceptic Emily Bender: ‘The emperor has no clothes’
Coding with LLMs
Some useful and practical tips on coding with LLMS:
- Refuse vibe coding most of the time. Instead rely on asking it questions and approving all the changes manually so you stay in command.
- Provide as much context as possible to LLMs because they thrive on it. Every time there's insufficient context passed on to them, they will struggle with their performance.
- Best models according to the author: Gemini 2.5 PRO and Claude Opus 4.0.
This reminded me of the tips by Paras Chopra but more specific to Cursor.
