I’ve started an AI-drafted blog
This one is about Bangalore weather
Karthik S on startups, meditation, job hunting, and personal reflections.
This one is about Bangalore weather
At a society and organisation level we’ve come to prioritise urgent work over important work. Which makes it incredibly hard to do well on important but not urgent work
I found India-made tote bags impossible to use, since their handles are way too short. No such issues with foreign totes
One prompt takes too long to get answered, and so you give another prompt in another project. Now that takes too long, and soon you’re not sleeping!
There is much greater returns to “intrapersonal vertical integration” now. A lot of the old specialisations and silos will go away since they make LLM-driven programming inefficient
Having learnt it the hard way recently, one thing I'm NOT going to outsource to AI is data exploration. And this is based on an idea from a 1977 book by John Tukey!
People will continue to remove "AI tells" from their writing. AI models will evolve to "write less like AI". New AI quirks appear, which people will distance from. And the cycle goes on...
They remind me of old (documented) thoughts, and can sometimes develop surprisingly good connections between them
In all my experiments I kept buying suboptimal coffee, so got Claude to build a recommendation system to help me shop better
With AI, it is easy to be abundant in your production, but the user's attention is still limited. So the creator's job is to curate, and direct the user on where to pay attention
I bring together Thinking Machines (the AI company), Databricks and Taaza Thindi (a South Bangalore restaurant) to talk about how open source can be a massive flex.
How you choose to write - with pen on paper, or on a computer, or use a scribe, or use voice typing - depends on your speed of thought generation for whatever you're looking to write.
Maybe because they are so verbose, only LLMs can read LLM-authored documents or code
Indians look down on "taking the easy way out" which is exactly what GLP-1s are seen as. And they are also "artificial, Western innovations", which doesn't help
What I used to call "Bayesian recognition" is, I realise, a special case of "human context rot". And we can use our understanding of LLMs to understand our own thinking better!
Heminway's way of working - where you know at the end of a workday precisely what you'll work on the next morning - can be an antidote to doing too much fighter work. But then there is Ganesha!
In 2 weeks' time, I'm doing a workshop on making awesome visualisations using LLM. Here is a nice quick exmaple
People say children are like an avatar of Vishnu. Among all avatars, the best candidate is Parashurama. Read on to know why.
Employees and consultants are wired differently. Employees excel at BAU work, while consultants always work "on the edge" (transforming stuff). Most people are wired for exactly one of these
Data science team succession is tricky when the models are used for a small number of high-stakes decisions. Did Liverpool get this wrong post Ian Graham?
The Takshashila Institution has made public a whole bunch of skills, based on their public policy research and writing over the years. I take a few for a spin.
What makes AI art art? When does AI become a tool in the hand of the artist, and not the creator itself? The answer lies in "one touch"!
Imam Pasand is very dense in sugars, and relatively low on fibre and water, and so it can give you a really massive sugar spike. It is no doubt sweet and tasty!
A very belated realisation that running a funded startup can quickly become a rat race!
There was a Project I wanted to work on using both ChatGPT and Claude. My "elegant" solution has been to keep the files on my computer, and use Claude Code and Codex instead!
Biceps have come to represent strength since they are the only part of the male body that can become muscular, and where such muscle can be shown off while wearing "polite clothing"!
Revisiting football club elo data to analyse the current state of top English clubs.
A curious thing about how I learn - whenever the first step of something comes to me easily, I find it really hard to learn. When I grind through initially, later on I ramp up really well`
Think of food in terms of 2 axes - good and bad for health, and enjoyable or not. Just cut the "bad for health and not enjoyable" quadrant (in my case that has a lot of density).
I was at IIMB for a panel discussion today. Some pertinent observations both from the panel and my visit to campus.