SK 🐵 @smilingArab

CTO @ Folio — changing the hiring paradigm, one match at a time. salehk.io Bay Area Joined September 2010

Tweets

3K
Followers

218
Following

371
Likes

798

dax @thdxr

3 months ago

LLMs are hard to create a moat around it's stateless compute that you can switch overnight when a better/cheaper option shows up all the commotion you see is downstream of this fact as companies flail around trying to fight this

107 95 2K 147K 220

View Details

SK 🐵 @smilingArab

6 months ago

@GergelyOrosz I just reached out to their support about this. I am on the $200 max plan and used 50% in 2days. Not sure how that’s possible.

0 0 1 859 0

View Details

SK 🐵 @smilingArab

7 months ago

I think I've finally found a way to solve the problem of memory in my daily workflows when working in Cursor. For any large piece of functionality or feature, have Opus create a PRD or Wiki. It may start small or large, but now whenever you revisit that functionality or feature, you reference the Wiki/PRD and use it as a living document that explains the history of changes and the expected behavior, so you never have to re-explain yourself again. I'm sure Cursor will soon add this as a feature to their app.

0 0 0 179 0

View Details

SK 🐵 @smilingArab

7 months ago

I figured out how to join the 10k lines-per-day club. Commit the PRD's and Cursor plans.

0 0 0 129 0

View Details

SK 🐵 @smilingArab

7 months ago

@aquariusacquah Wow. That’s a lot of Slop!

0 0 0 9 0

View Details

SK 🐵 @smilingArab

7 months ago

Everyone keeps saying SaaS is dead. AI killed SaaS. But we’re all using Cursor and Claude Code which is SaaS

1 0 0 128 0

View Details

SK 🐵 @smilingArab

7 months ago

These statements are pretty wild because anyone who actually builds anything functional on a daily basis realizes how crazy 10k lines per day. You’re telling me you added 50k lines of code to your code base in one week? What are you possibly building or creating that requires that many lines of code in that short amount of time? If you tell me you’re writing 10k lines of code a day but 7k of them are unit, integration and e2e tests I might slap you in the face.

5 1 43 8K 1

View Details

SK 🐵 @smilingArab

7 months ago

@signulll Software and product quality is bad.

0 1 0 305 0

View Details

SK 🐵 @smilingArab

7 months ago

@Angaisb_ How is Claude code different than using Cursor with Opus4.5?

1 0 0 1K 0

View Details

SK 🐵 @smilingArab

8 months ago

@rauchg I’m currently upgrading to Nextjs 16 and I have encountered a very unique bug in dev mode. How can I get in touch with someone on the Nextjs team to get support on this. It’s blocking me from being able to deploy the upgrade to prod. I have a work around but it makes no sense why it would work.

0 0 0 14 0

View Details

SK 🐵 @smilingArab

9 months ago

Coding Agent fatigue hit me today. I stared at my screen thinking, I don’t feel like describing what I want to do today. I’m just going to do it.

0 0 0 68 0

View Details

SK 🐵 @smilingArab

9 months ago

@reactjs Sounds like RSC was a bad design choice.

0 0 1 257 0

View Details

SK 🐵 @smilingArab

9 months ago

@dbreunig Microsoft software hasn’t been good in a long time. They’re lucky that they have a bunch of enterprises locked into their software. I don’t think anyone can name a notable product that Microsoft has released in the past decade.

1 0 2 187 0

View Details

SK 🐵 @smilingArab

9 months ago

@flaviocopes They can practice by not using AI, or using AI to teach them instead of giving them answers.

0 0 0 22 0

View Details

SK 🐵 @smilingArab

9 months ago

@garrytan I don’t think OpenAI gives a better answer for a much longer wait. Gemini 3 and Claude 4.5 are the best and most consistent models right now. I work with them both interchangeably and him them check each others work.

0 0 1 94 0

View Details

SK 🐵 @smilingArab

9 months ago

The day an agent or LLM can take an ambiguous prompt or PRD from someone who never learned to code and build a long-lasting, properly scaling product is still a long way off.

0 0 0 41 0

View Details

SK 🐵 @smilingArab

9 months ago

It all comes full circle: without solid foundations and fundamentals, you’ll eventually hit a wall where the agent can’t progress, and you’ll create more problems than you solve.

1 0 0 68 0

View Details

SK 🐵 @smilingArab

9 months ago

My honest opinion on the “SWE’s are cooked” take: I don’t think it will happen anytime soon. Coding agents are far more useful to me because of the experience I’ve built over the past decade.