Mher Movsisyan

Builder by nature. Software Engineer by specialty. I use code to solve problems and curiosity to find them.

[Mher Movsisyan @MherMovsis]

@alexandr_wang I can’t download it

[Mher Movsisyan @MherMovsis]

@alexandr_wang Is app available in Canadian App Store?

[Mher Movsisyan @MherMovsis]

@pitdesi Especially on the UI implementation part

[Mher Movsisyan @MherMovsis]

@iakob44058343 @tobi That's strange! When I press Download link on the https://t.co/mrray7TMur webpage it navigates to the app store and shows App Not Available.

[Mher Movsisyan @MherMovsis]

@pitdesi It’s not available in the Canadian App Store

[Mher Movsisyan @MherMovsis]

@finkd @AIatMeta Will check it out

[Mher Movsisyan @MherMovsis]

To survive personal builds, you end up model-routing: juggling high efficiency options like DeepSeek V4 Flash or Meta Muse Spark 1.3 for grunt work, then swapping back to heavyweights for edge cases.

Where the compute budget exists, the blast radius forbids it. Where autonomy thrives, the API limits pull the brake.

[Mher Movsisyan @MherMovsis]

At a corporate job: You have near-infinite tokens. But you’re dealing with a 10-year-old legacy codebase used by millions. You can’t just unleash 10 autonomous agents without blowing things up. Human in theloop and risk management keep your rate of change low.

[Mher Movsisyan @MherMovsis]

On personal/MVP projects: The repo is greenfield, context ready, and has zero legacy baggage. You can map the architecture, launch 5 parallel sub-agents across git worktrees, and watch features build simultaneously.
The catch? Your plan runs dry in 35 minutes.

[Mher Movsisyan @MherMovsis]

There are two completely opposite realities of coding with AI right now:

• Corporate: Massive compute budget, strictly limited velocity • Personal: Insane velocity, hit the token wall in 40 minutes
A quick breakdown 🧵

[Mher Movsisyan @MherMovsis]

@DaveShapi Using with API credits? With subscription on 4 concurrent projects you would hit a limit in 5-10 minutes.

[Mher Movsisyan @MherMovsis]

@tobi @tokifyi Do it and I would love to join and work on it

[Mher Movsisyan @MherMovsis]

@elshayib_ What’s the source of this photo?

[Mher Movsisyan @MherMovsis]

@Henryf1w No, it's pretty good! Haven't done much task implementation as it hit token limit very fast, but used it to review some critical workflows and identify possible improvements.

[Mher Movsisyan @MherMovsis]

@BlakeWhittle7 I just use it to give general improvement strategy for some features, then lunch a lot of other subagents to implement the suggestions that make sense for me.

[Mher Movsisyan @MherMovsis]

@kapilansh_twt I was thinking to install it back and give another try

[Mher Movsisyan @MherMovsis]

@lydiahallie Thanks!

[Mher Movsisyan @MherMovsis]

@pitdesi Most of AI wrapper startups are falling one after another

[Mher Movsisyan @MherMovsis]

@petergyang Hit the limit 3 times in last 24 hours and it usually takes 30-45 minutes of agentic coding to get to the limit.