Harinder Takhar

6.8K posts

Harinder Takhar banner
Harinder Takhar

Harinder Takhar

@makingmusic

तत्त्वमसि, realizing it in this life through paytm, paypay and woodworking.

Mississauga, Ontario Katılım Temmuz 2009
640 Takip Edilen3.2K Takipçiler
Harinder Takhar
Harinder Takhar@makingmusic·
And so will the meaning of shareholder value.
English
0
0
0
39
Harinder Takhar
Harinder Takhar@makingmusic·
Everybody who has ever built software is experimenting at an unprecedented level and the best ones amongst us are pushing the limits to understand how the world will work tomorrow. This is fun @deedydas thanks.
Deedy@deedydas

Karpathy's Autoresearch pushed my vibecoded Rust chess engine AI from "expert" to a top 50 grandmaster, a #311 chess engine. It ran over 70 experiments on its own and tried to hill climb to the top ELO score it could, landing at 2718!

English
0
0
0
604
Harinder Takhar
Harinder Takhar@makingmusic·
on the other extreme there are systems that are 50% good 100% of the time. those are dead on arrival. you have to get out of that zone.
English
0
0
0
49
Harinder Takhar
Harinder Takhar@makingmusic·
people make the mistake of making a system that might be 100% good, but only 20% of the time. and try to work on that 20%.
English
1
0
0
51
Harinder Takhar
Harinder Takhar@makingmusic·
if you can build a system that is 90% good 100% of the time, you are there.
English
1
0
0
93
Harinder Takhar
Harinder Takhar@makingmusic·
the future is not "which agent is the best". the future is "who controls the workflow"
English
1
0
2
77
Harinder Takhar
Harinder Takhar@makingmusic·
@garrytan @danshipper And the best parallel-izers are not the best programmers or even the best product managers. It's a completely new skill.
English
0
0
0
30
Harinder Takhar
Harinder Takhar@makingmusic·
The maturity of using parallel agents is limited by humans. Already. People who will learn parallelization of tasks better than others will get an order of magnitude more speed than the others. @garrytan @danshipper are me heroes in this.
Dan Shipper 📧@danshipper

prompt get many PRs to prod autonomously using codex subagents: Run a continuous prod-to-green swarm loop. Keep the immediate blocking task local. Use a small stable set of persistent subagent lanes: 1. prod monitor 2. staging shepherd 3. current/newest pathology investigator 4. current fix worker owning the patch/worktree Manage subagents actively: - Give each agent one durable role, one owner lane, and one concrete output contract. - Reuse agents with send_input when new evidence appears; do not respawn unless the lane is genuinely new or the old agent is stuck. - Treat new information as first-class work: when the main thread or another agent learns something material, decide explicitly which existing agent should receive that delta. - Ask agents to report in a compact stateful format: current belief, what changed, confidence, next action, blocker if any. - Require monitors to stay persistent and report only on meaningful state changes, not one-shot summaries. - Do not close or interrupt agents casually; only do it when the lane is complete, superseded, or clearly mis-scoped. - Prefer fork_context=false for narrow review/monitoring tasks; use fork_context=true only when continuity from prior lane context is actually needed. - Poll sparingly. Wait only when blocked on that agent’s result. For every delegated task, require concrete outputs only: - evidence - likely root cause - smallest failing test - smallest safe fix - focused validation - commit SHA if code changed - residual risk - whether this creates NEW_PATHOLOGY or is same-family noise If NEW_PATHOLOGY appears, keep existing monitor lanes running and spin one fresh investigator + one fresh fix-worker lane for that pathology. Optimize for the fastest safe path to prod green. Keep going until prod to green

English
1
0
0
187
Harinder Takhar
Harinder Takhar@makingmusic·
Macro hard must be harding right now.
Andrej Karpathy@karpathy

@ChristosTzamos Wait this is so awesome!! Both 1) the C compiler to LLM weights and 2) the logarithmic complexity hard-max attention and its potential generalizations. Inspiring!

English
0
0
0
102
Harinder Takhar
Harinder Takhar@makingmusic·
What @garrytan is doing is the top 0.01% of ai-pilled people are doing. He is orders of magnitude ahead. The ability to parallelize your tasks is under rated.
Garry Tan@garrytan

I'm going to rile up the trolls with this right now but I am working on 3 different big projects simultaneously across 15 @conductor_build sessions all the time. In the last 7 days I'm averaging 17k lines of code per day, 35% tests, all thanks to gstack. (All mornings/nights/weekends on top of my real job at YC) I ran /retro (from gstack) on all three projects and this is what came back:

English
0
0
3
613
Harinder Takhar retweetledi
matt rothenberg
matt rothenberg@mattrothenberg·
just picked up this bad boy. can't wait to write some software with it
matt rothenberg tweet media
English
223
881
15K
595.3K
Harinder Takhar
Harinder Takhar@makingmusic·
TIL about the /btw command in claude code
English
0
0
1
347