Yash Khivasara

456 posts

Yash Khivasara

Yash Khivasara

@k2_yash

building AI tools at 2am because the idea wouldn't shut up. 3 products shipped. more coming.

India Katılım Mart 2021
469 Takip Edilen138 Takipçiler
Sabitlenmiş Tweet
Yash Khivasara
Yash Khivasara@k2_yash·
i'm yash. i build AI tools because i can't help it. here's what's live right now: growero.io — AI that manages your social media so you don't have to rexels.app — AI YouTube thumbnails that actually get clicks rivsy.app — blogs optimized for Google AND ChatGPT all built nights & weekends while working full-time as a data engineer. if you're building something too, let's connect.
English
0
0
3
210
Yash Khivasara retweetledi
Tyler
Tyler@rezoundous·
Microsoft canceled Claude Code license due to unsustainable costs. If they can't afford it, who can?
English
348
1K
12.7K
722.8K
Yash Khivasara
Yash Khivasara@k2_yash·
spent 20 minutes telling claude code what NOT to do. it listened perfectly. then did exactly the thing i told it not to do, but in a different file.
English
0
0
0
155
Yash Khivasara
Yash Khivasara@k2_yash·
@adisingh @theo which tasks? for one-shot stuff, yeah codex wins. for anything multi-file cc still handles context way better. annoying rules but i haven't found anything cleaner for actual projects.
English
1
0
0
185
Adi Singh
Adi Singh@adisingh·
I saw @theo's video about ditching Claude Code and thought "it couldn't be that bad, right?" It is that bad. Codex is so much better.
English
15
8
228
21K
Richard | £1M Journey 🇬🇧
In my lifetime I can think of 2 key technology pivot points that signaled a change to everything. 1. Personal Computer 2. iPhone I have recently experienced the 3rd and most people haven't even used them yet. AI agents You need to get one and start learning...NOW
English
6
0
19
1.5K
Yash Khivasara
Yash Khivasara@k2_yash·
@IAmAaronWill the skip approach makes sense for transient stuff. where i'd add a layer is error rate tracking. if 30 out of 100 leads are skipping, that's not a lead problem. that's a scraper problem. do you log those?
English
0
0
0
1
Aaron
Aaron@IAmAaronWill·
@k2_yash Error handling sub agents. And a workflow that still continues if there's an error. My lead scraper for example. If a lead gets scraped but it comes back as an error for whatever reason, the system skips that lead and moves on instead of breaking.
English
1
0
1
35
Yash Khivasara
Yash Khivasara@k2_yash·
@IAmAaronWill works great until one breaks quietly. n8n dies at 2am and the other 4 don't notice. you end up being the monitoring layer for all of them. how do you handle that part?
English
1
0
0
43
Yash Khivasara
Yash Khivasara@k2_yash·
claude code keeps writing tests that pass by asserting the bug instead of the fix. green checkmark, feature still broken. i've started reading the tests before i trust the green.
English
1
0
0
29
Yash Khivasara
Yash Khivasara@k2_yash·
@AzFlin the one place step 1 breaks for me is when the model needs data it was never trained on. that's the only reason i reach for retrieval. everything past that is usually solving boredom, not problems.
English
1
0
1
33
AzFlin 🌎
AzFlin 🌎@AzFlin·
Same thing applies in AI > use this LLM > wait it’s outdated, use this new one > now you need long term memory setup ofc, your AI needs to remember things > ok now you need this agentic harness, it’ll supercharge your LLM’s capabilities > but you’re using too many tokens now, try this token reduction skill > you’re only running one agent? What a scrub, you need to parallelize. Use this multi agent orchestration tool. And on and on and on. To infinity. And guess what .. for most people, the most effective path would have been to stop at step 1 Just use this LLM.
GIF
Aakash Gupta@aakashgupta

Every certified personal trainer alive would look at this program and call it garbage. No progressive overload, no rest days, no pulling movements, no periodization, no protein timing. Three exercises and running. That's it. For 1,095 straight days. Tasuke got more shredded than 95% of people paying $200/month for optimized coaching. Run the bull case for modern programming. Periodized splits with progressive overload produce faster hypertrophy in controlled studies. Undulating rep schemes prevent plateaus. Pull/push balance prevents injury. Rest days allow supercompensation. The science is real and the results are measurable. A good coach will get you further in 12 weeks than Saitama's routine will in 12 weeks. No question. Sounds like a win for complexity until you realize what Tasuke actually traded it for. He traded optimization for the one variable that beats all of them: a program so simple he never had to think about whether to do it. 100 push-ups. 100 sit-ups. 100 squats. 10km. Go. The $30 billion fitness industry sells periodization, app subscriptions, macro calculators, and recovery protocols because those are renewable revenue. Adherence to three exercises for 1,095 days generates zero recurring fees. You can't monetize "just keep showing up." 328,800 total reps. 10,000 miles. Built on a routine a fictional bald superhero made up as a joke. The program was never optimal. The program was never supposed to be optimal. The program was supposed to be impossible to talk yourself out of on a Tuesday morning. And that turns out to be the only fitness variable that compounds.

English
9
1
24
3K
Yash Khivasara
Yash Khivasara@k2_yash·
@rozzabuilds agree mostly. but codex still wins for one-shot scripts for me. claude code fixes the bug then refactors 6 files i never touched. what are you keeping the $20 codex around for?
English
0
0
0
55
Rozzabuilds
Rozzabuilds@rozzabuilds·
Unpopular opinion Claude Code > Codex
Rozzabuilds tweet mediaRozzabuilds tweet media
Català
11
0
15
621
Yash Khivasara
Yash Khivasara@k2_yash·
claude code can ship working code at 11pm. it still can't tell me if the feature should exist. that part is still depressingly mine.
English
0
0
0
12
Yash Khivasara
Yash Khivasara@k2_yash·
@plainionist kinda already exists. CLAUDE.md + slash commands + skill descriptions. we just keep calling it markdown.
English
0
0
1
11
Seb
Seb@plainionist·
Serious question: How long until Markdown is replaced by a DSL optimized for instructing AI agents? 😉
English
38
1
25
4K
Yash Khivasara
Yash Khivasara@k2_yash·
@aryanlabde question assumes claude/codex are doing the coding. for me they're doing the typing. the thinking part hasn't moved.
English
0
0
0
27
Aryan
Aryan@aryanlabde·
Vibe coders, can you really code without claude and codex?
English
53
1
36
1.9K
Yash Khivasara
Yash Khivasara@k2_yash·
@zamdoteth writing the dms is the easy part. agent doesn't know which 100 humans would actually care. most of us don't either when we ship.
English
0
0
0
131
zam
zam@zamdoteth·
God i wish there was a claude code/codex for getting users to an app Hey claude, get me my first 100 users Who’s building this??
English
32
2
59
5.6K
Marcus
Marcus@MarcusSpillane·
@k2_yash @swyx My favorite genre is when the migration guide is longer than the code you're migrating. Three pages of instructions to change an import path. You know someone got a sprint's worth of credit for that rename.
English
2
0
0
14
swyx
swyx@swyx·
this is a big deal, on the order of Kelsey Hightower’s “Kubernetes The Hard Way” and probably all ai engineers should go thru this once mostly i advocate “just in time learning”, but this is one scenario you want “just in case”
swyx tweet media
Ahmad@TheAhmadOsman

x.com/i/article/2050…

English
39
42
669
191.6K
Yash Khivasara
Yash Khivasara@k2_yash·
testing my voice agent by calling it and pretending to be a borrower who forgot about their payment. i'm now a better debt collector than the bot. not sure how to feel about that.
English
0
0
0
8