27th July, 2026
Updated: Aug 17
What am I working on?
On another fun creative. Buy the t-shirt here
What am I reading?
I read another masterpiece called The Arrival by Shaun Tan. This is a wordless, surreal graphic novel that shows journey of an immigrant (probably fleeing atrocities in his homeland) to start anew in an unknown land. The soft charcoal like illustrations, ranging from neatly organized square panels to sweeping full page spreads has such depth without the use of a single word. It captures the disorientated feeling the immigrant feels upon arrival in a completely unfamiliar world: strange architecture, a new language, bizarre customs, and unearthly creatures (although they are friendly). It grounds this surreal setting in the everyday challenges of survival - finding shelter, earning a livelihood, navigating public transit or finding food.
I like the sequences of cloud (to me it depicted the sky on consecutive days of his journey to the new land), the surreal illustrations of the new land, the moment when he opens his briefcase and the image shows his wife and daughter eating without him, and wordless backstories shared by fellow immigrants revealing why they, too, had to flee. This book can be interpreted in your own way and that is the beauty of it. It is a MASTERPIECE and the more I read such books, I continue to be amazed at the expansion power of this medium. To me reading this book felt like being a child again who doesn't know how to read and makes sense of the story by looking at the visuals and their imagination.
What am I watching/watched recently?
I watched Mary and Max last weekend. It a beautifully crafted stop-motion animation that explores the unlikely pen-pal friendship between Mary (a young girl in Australia) and Max (a middle-aged man living in New York). Inspired by a real life correspondence of the director, the film begins with Mary’s whimsical letter to a stranger in the U.S. Through letters, gifts, and shared stories, the two find in each other the companionship they had long been missing in their lives.
The film’s aesthetic is unique: clay characters set against predominantly grey and muted backdrops with the narration unfolding almost entirely through letters sent to each other. However, the movie deeply explores serious topics like anxiety, depression, loneliness, and social alienation. Max eventually discovers he suffers from a mental health condition. While the friendship drifts apart (due to some situations that I would not reveal) the two ultimately reconnect.
Despite its artist value, I found the film overwhelmingly melancholic. Themes of absentee parenting, broken relationships, suicidal thoughts, and isolation weigh heavily on the storyline. The use of dark humor makes some scenes funny but it is far from lighthearted (certainly not a film for children). Overall, for me it felt more like a downer than a delight.
New thing I learnt lately?
The 500 denomination note is the most widely circulated note in the country, followed by the ₹10 note. Our notes are are made from a special kind of paper that’s made from a blend of 75% cotton and 25% linen. This paper can withstand ~4,000 folds before it finally tears.
Tokenomics is what I learned this week. Tokens have become the fundamental economic unit of AI. Tokenomics (AI Token Economics) is the discipline of measuring how tokens are consumed and linking that consumption to business cost and benefit.
Token are of two types:
Input tokens = prompts, instructions, retrieved context, conversation history
Output tokens = generated responses, tool calls, reasoning traces
Some interesting things I came across:
Price per token is falling with increasing efficiency in model, but the overall token consumption is increasing (Deploying AI agents, Running multi-step workflows, Generating images, audio, and video, Adding reasoning models)
Traditional costing models with cost/user license is moving to seat-fee-plus-pre-committed-token-consumption structure. Costs can be non-linear making governance critical.
It requires the metering of token quality alongside token quantity. An organization that tracks only volume will misattribute cost. Therefore, model tiering (routing simple tasks to smaller models, reserving frontier capability for the queries that require it) so that results are achieved accurately and at required speed as per the set objective. This leads to a concept called 'Goodput' in AI.
How can the cost be optimised? Couple of ways - Context compression and pruning, using newer formats designed specifically for LLM consumption, model routing and code execution (as models are stronger at writing code that calls tools than at selecting tools through a predefined set)
Comments