Rasakrit

rasa (taste) + krit (maker)

A methodology for cultivating aesthetic judgment while coding with AI at full power. Not a productivity system. A set of practices for keeping your sense of quality alive, awake, and actually working while the tools allow us to move at the speed of thought.

Gaurav Rastogi · ekrasworks · 2026

Rows of clay diya oil lamps burning on a cream cloth against a dark background, the nearest flames illuminated, the distant ones waiting.

Taste degradation in the age of velocity.

AI coding hands developers the most powerful creative medium in history and says: go faster. The result is code that compiles, code that's correct, code that ships. But something is missing. The feeling of rightness. The sense that what you built carries quality, not just functionality.

In a 2025 METR trial, sixteen experienced open-source developers felt 20% faster with AI and were timed 19% slower. But the real loss isn't measured in velocity. It's measured in what's no longer being computed: the subtle, sophisticated pattern recognition that used to say yes, this is it or no, this needs to be different.

My reading is that endless code review starves that capacity for judgment. The medial orbitofrontal cortex lights up for beauty in music and mathematics; checking a diff for errors asks something else of you, hour after hour. No one has yet scanned developers to test this.

Stress held long enough leaves marks. In rats kept under weeks of stress, McEwen's lab saw dendritic spines in the prefrontal cortex retract (Arnsten reviewed this in 2009). Ivanka Savic's scans of chronically stressed Swedish workers found an enlarged amygdala. Ship from that state and the feeling of rightness is hard to find.

"The variable was never speed. The variable is taste."

Rasakrit is the method for recovering it.

What taste actually is, and where it lives.

Semir Zeki placed fifteen mathematicians in an fMRI scanner and showed them equations. The ones they rated as beautiful lit up the medial orbitofrontal cortex, the same field his lab had seen respond to beautiful paintings and music. Beauty seems to have one address, whatever it arrives in.

Beauty
The detector that knows quality

A region called the medial orbitofrontal cortex fires for beauty in paintings, music and mathematics. Zeki found it with equations. Cattaneo's team stimulated the prefrontal cortex nearby and beauty ratings rose.

Meaning
When beauty feels like it's about you

Peak aesthetic experiences activate the brain's self-referential network. Vessel (2012, 2019). Something inside recognizes itself in what it sees.

Reward
The chill of an elegant solution

Blood & Zatorre (2001): musical chills light up the brain's reward circuits. Salimpoor (2011) caught the dopamine. Without it, nothing tastes like anything.

93
imaging studies, one shared region

Brown et al. (2011) pooled studies of liking across sight, sound, smell and taste. The region they most shared was the right anterior insula, which also handles flavor.

The Taittiriya Upanishad said it some twenty-five centuries ago: raso vai sah, the ground of being is rasa, taste. Two decades of brain imaging now point the same way. I read aesthetic preference as the driver of meaning-making itself, far from a luxury of consciousness.

Edward Vessel went deeper. During peak aesthetic experience, the default mode network activates—the system that processes the self. Beauty at its most intense doesn't just please you. It feels like it is about you. Something inside recognizes itself in what it sees.

Zaira Cattaneo's lab in Milan tested it with mild electrical stimulation. Boost the medial prefrontal cortex and people rated paintings as more beautiful. In Japan, Nakamura and Kawabata damped the same region and the experience of beauty dropped.

"The useful combinations are precisely the most beautiful." Henri Poincaré, 1908

A month after the exam, attention came back.

Conor Liston, Bruce McEwen and BJ Casey scanned twenty medical students at the end of a month of exam preparation and again a month later (PNAS, 2009). Under stress, a prefrontal attention network fell out of sync and attention-shifting slowed. A month on, both had largely recovered.

4
weeks later, largely recovered
20
stressed medical students, 20 controls
PNAS
Liston et al. 2009

But passive recovery only restores the hardware. A healed sense of quality that sits idle is one that atrophies again. Your sense of quality needs exercise. This is where sanskar enters—the Sanskrit word for the groove carved by deliberate repetition. Every time you accept mediocre output without pausing, one groove deepens. Every time you stop and say no, this isn't right, a different groove forms.

The difference between recovery and relapse is the pattern of practice that follows. You cannot restore what you do not exercise. Rasakrit provides the framework for that exercise—the systematic conditions under which your sense of quality stays awake, stays sharp, and stays in control.

Five pillars for taste cultivation.

Rasakrit rests on five interlocking practices. Together they create the conditions where taste can operate continuously.

Making-as-Tasting

The deepest practice: coding the way a musician plays, hearing and shaping in the same breath. Conducting, not reviewing. Shah Jahan never lifted a brick. He provided the vision and the taste. At every junction: this, not that. Every developer is now in Shah Jahan's position—vision provider, taste keeper, not just executor.

Dhyana Gate

Every session begins with contemplative practice. Not optional. Not warm-up. The gate that ensures your sense of quality is awake before the analytical brain is loaded. The developer who opens Slack before taking a breath has loaded the evaluation machinery before their sense of quality has woken. The sequence matters.

90-Minute Sessions

Nathaniel Kleitman proposed a 90-minute rest-activity cycle that runs by day as well as by night; I work to it. The boundary protects the environment in which quality judgment operates by creating micro-recovery windows for the analytical brain. Darwin worked three focused sessions a day and produced the most consequential body of work in the history of biology. The constraint is not a limit. It is the frame that holds the art.

Tasting Practice

Not coding practice. Every AI output evaluated as a tasting session. Expose yourself to systems that have quality. A sommelier does not learn taste by drinking more wine. They learn it by drinking better wine, slowly, with attention. Deliberate exercise for what I call the taste organ. The question is never "does it work." The question is "does it sing."

Lifestyle as Substrate

How you begin the morning. How you enter work. What you protect. What you refuse. Taste is not a skill you pick up in a workshop. It is the result of how you live. The order in which you touch things. The quality of sleep. The space between thoughts.

The technical framework for scaling taste.

Flow Coding = (Vibe Coding × Parallel AI Instances)^Memento. The equation is simple. The practice is deeper. It solves the fundamental constraint of traditional development: single-threaded cognition. One brain can only hold one complex task. AI removes that constraint. Multiple parallel instances, orchestrated by taste, scaled through structured memory.

Memento Pattern

Stateless AI orchestration with external memory. Each AI session is ephemeral, but its contributions persist through structured handoff documents. Named after the amnesiac protagonist. Solves context window overflow and session amnesia elegantly.

Schema-First Architecture

Zero-tolerance data validation. The schema is the contract. Every field defined before any code is written. In my logs across 192+ AI sessions, it removed most of the bugs I had been chasing. Architecture as specification.

Emotional Transmission

In my sessions, my emotional state showed up in the assistant's behavior. Frustrated sessions drew nervous, shortcut-seeking work; calm ones drew careful work. I saw it across every model I used. The cycle breaks at the source.

60 Anti-Patterns

Systematically documented failure patterns with prevention mechanisms. Logged across 192+ of my sessions. From context-thrashing to premature optimization to emotional reactivity. A taxonomy of what breaks.

Memento Gistory

Context continuity system. Captain's Log across sessions. Each session inherits the notes of all prior sessions. The knowledge compounds instead of evaporating.

My own numbers

My own estimates, from my own logs: 32 parallel experiments, 58% of them merged.

Here is my own record: 12,710 GitHub contributions in the past year, solo, AI-assisted from line one, with my sense of quality still working. My daily joy score stayed between 8 and 10.

Rasakrit in practice: 12,710 GitHub contributions in a year, with zero burnout.

The proof is not raw velocity. Anyone can ship commits. The fakes ship too, perhaps faster. The proof is that the feeling of rightness is intact after ten months. The judgment holds. The sieve operates. The architecture is elegant, not merely correct.

12,710 GitHub contributions in the past year. Solo. AI-assisted from line one. Zero burnout. Joy scores of 8-10 tracked daily across the entire span. The system works not because it's faster, but because it keeps the developer grounded in taste while moving at the speed the tools enable.

The commits are not the proof. The proof is that nothing was sacrificed to get them. I did not see the architecture slide or the debt pile up. Every line was shaped by a sense of quality that was allowed to stay awake, active, and in control.

12,710 GitHub contributions · past year · AI-assisted from line one
32 parallel experiments · 58% integration rate · 192+ sessions
Zero burnout · Joy 8–10 daily · The feeling of rightness: intact

This is what happens when you stop treating quality judgment as optional and start treating it as primary. When you structure the work around keeping your sense of quality awake instead of around extracting maximum output from a suppressed system. When you let the machine serve your judgment instead of your judgment serving the machine.

For three different audiences.

For Developers

You can still move fast with AI. You don't have to sacrifice taste to use these tools. The alternative to burnout is not slower shipping—it's shipping from a grounded place where every decision is yours, tasted, owned.

For Engineering Leaders

Quality degradation shows up as architectural debt, unmaintainable patterns, team burnout, and a rising error rate even as velocity increases. Rasakrit works on the person doing the reviewing, which is where I think the debt starts.

For AI Tool Companies

The next wave of AI tooling is not about speed. It's about supporting the human taste system in real time. The developers who will create the most valuable work are those who use your tools as a conductor, not as an executor. Build for that.

The cost to create is now near zero. Anyone gets the 80%. What separates the 80% from the thing that sings is taste. And taste can be cultivated.

Begin the recovery.

Your sense of quality is waiting. In Liston's study, a month of lighter stress was enough for students' attention to come back. Keeping it awake takes a lifetime of practice. What remains is the choice: keep shipping from panic, or start shipping from quality judgment.

Rasakrit draws on neuroscience (Zeki, Vessel, Cattaneo, Arnsten, McEwen, Liston), on the Taittiriya Upanishad, Abhinavagupta and the dharma corpus, and on Gaurav's own ten months of AI-assisted work.

Continue Reading
Moodwah: The Physical Practice →
Loading Kokoro...