HUNAR Jain
ISSUE No. 47 · BAY AREA · VOL. MMXXVI
← back to the writing desk
ESSAY · 2026-10-11 · 4 min · learning · - views

Steering Without a Map

A working model for one person learning in a world that keeps changing while they learn it.

by Hunar Jain · ~4 min read

You’re an agent learning in a world that won’t hold still. The rules drift while you act, outcomes are noisy, and you only ever see part of the picture. You can’t solve it. You can only steer.

Fig. 1 · The goal drifts and the currents shift while you move. You never solve the path. You keep correcting it.

Two kinds of cues

Reality talks back in two ways. Mechanical cues come from things that don’t adapt to you: your skills, your code, your body in the gym. They’re slow to read, but what you learn sticks. Social cues come from people who model you back: managers, peers, markets. They react to you, and they keep changing.

Fig. 2 · Mechanical systems repeat the same rules. Social systems turn to look at whoever is learning them.

Let reality show you where you’re slow

Everyone adapts fast in some areas and slowly in others. You don’t need a test to find out which. Set a goal, act, and compare what you aimed for with what happened. The gap is the signal. Show up late to meetings three times and reality has told you something about your time management. Adjust, and the gap closes. Keep missing, and you’ve found a place where you need more deliberate reps.

Fig. 3 · The gap between aim and outcome is the measurement. Three misses in the same direction mean adjust.

Shrink the social target

Social systems drift faster than you can learn them, so don’t try to learn everyone. Pick the narrow, stable set of people you’ll be around for a while and who actually move your outcomes. Learn them, and watch how they change.

Fig. 4 · The crowd moves too fast to model. A few people who stay put long enough are learnable.

Update everything at once

When you miss a goal like a promotion, it’s tempting to ask whether it was the work or how the work was seen. Usually you can’t tell, and you don’t need to. Improve the skill and read the right people at the same time. Let reality grade the whole update over a few quarters. You’re not hunting for a single cause. You’re shifting weights everywhere.

Fig. 5 · Skill weights and people weights shift together. Reality scores the combined update each quarter.

What carries forward

Mechanical skill transfers fully and never leaves. Social learning transfers partially, as a rough model of how certain kinds of people behave. In a new setting, you look for the matches. That model is your biggest asset when you move, and your biggest risk: match too fast and you’ll misread people who don’t fit the old pattern.

Fig. 6 · Skill arrives intact. People patterns arrive as guesses: two match, one looks familiar and isn’t.

The shape of it

The loop is slow, noisy, and tangled. Judge yourself by trajectory, keep the loop running, and let reality do the teaching.

One miss is noise. The same miss three times is a message.

Fig. 7 · Single points jump around. The trend line is the thing to judge.

Still open

When do you trust the pattern you carried in, and when do you throw it out and read the new people fresh?

Read next

essay · 2026-10-07 · 1 min

When Output Stops Measuring Agency

Text and code are nearly free now, so output no longer works as a proxy for high agency, and the real thing is harder to tell apart from the performance.
essay · 2026-05-21 · 2 min

on power and the only invariants that overrule it

power gives you leverage to Exploit, Shun Noise, Control the narrative, Extra Perks like Money. but Death and class-agnostic suffering are the only invariants no one can keep their power through.