Mohammed S.

Indie developer + AI engineer. Building tiny AI tools and sharing the receipts 💻

[Mohammed S. @foundmod]

Glad Codex reset the limit, I was about to run out of limit today and wait for 3 days without any agentic work.

[Mohammed S. @foundmod]

@J_dev363 @rezoundous That’s why, the problem is the 5x plan

[Mohammed S. @foundmod]

Codex usage has been brutal this week, only 40% remaining and limit resets in 3 days.

with no /fast neither /goal and mostly medium/low.

Are you experiencing the same

[Mohammed S. @foundmod]

@J_dev363 @rezoundous Seems you are on the 200 plan

[Mohammed S. @foundmod]

@rezoundous Beo same here, no /goal no /fast that’s crazy

[Mohammed S. @foundmod]

@T_Zahil Same question, couldn’t find answer

[Mohammed S. @foundmod]

Small update on my TikTok marketing test:

I learned that setup can quietly become procrastination.

It feels productive because you’re configuring things, checking things, fixing small issues.

But the real learning starts when you study the market and test content.

Also, random VPNs seem like a shortcut, but they can create more confusion than progress.

I’m done tweaking for now.

Next step: observe, warm up, test, document.

Let’s see where this goes.

[Mohammed S. @foundmod]

I’m finally breaking into TikTok marketing after hesitating for way too long.

Today I finished my setup for testing TikTok content targeting the US market.

My setup:

- iPhone 7: $60
- IPRoyal ISP proxy: $10 for 90 days
- Shadowrocket: $2.99

Total cost: around $73.

The phone is dedicated only for this TikTok test.

I’m using one ISP dedicated proxy, one account, and keeping the setup simple for now.

The proxy is from IPRoyal:
- United States location
- SOCKS5
- routed through Shadowrocket on iOS

Before buying Shadowrocket, I tried free apps like V2Box and Streisand, but they were annoying for this setup.

Shadowrocket was cleaner.

One note: you may need a US App Store/payment method to buy it. I managed to pay with PayPal, but another option is buying a US App Store gift card.

I also don’t think using random VPNs is a good idea for this.

They can be detected by tiktok, overused, and create weird account issues. Then you spend time debugging the setup instead of learning what content works.

Today I created the TikTok account and warmed it up for 10 minutes.

No link yet.
Not posting immediately.
Just account creation and warmup today.

I’ll keep warming it up daily, watching niche-related content, lightly engaging, and documenting what works or breaks.

This is an experiment, not a guaranteed playbook.

Let’s see where this goes.

[Mohammed S. @foundmod]

I’m finally breaking into TikTok marketing after hesitating for way too long.

Today I finished my setup for testing TikTok content targeting the US market.

My setup:

- iPhone 7: $60
- IPRoyal ISP proxy: $10 for 90 days
- Shadowrocket: $2.99

Total cost: around $73.

The phone is dedicated only for this TikTok test.

I’m using one ISP dedicated proxy, one account, and keeping the setup simple for now.

The proxy is from IPRoyal:
- United States location
- SOCKS5
- routed through Shadowrocket on iOS

Before buying Shadowrocket, I tried free apps like V2Box and Streisand, but they were annoying for this setup.

Shadowrocket was cleaner.

One note: you may need a US App Store/payment method to buy it. I managed to pay with PayPal, but another option is buying a US App Store gift card.

I also don’t think using random VPNs is a good idea for this.

They can be detected by tiktok, overused, and create weird account issues. Then you spend time debugging the setup instead of learning what content works.

Today I created the TikTok account and warmed it up for 10 minutes.

No link yet.
Not posting immediately.
Just account creation and warmup today.

I’ll keep warming it up daily, watching niche-related content, lightly engaging, and documenting what works or breaks.

This is an experiment, not a guaranteed playbook.

Let’s see where this goes.

[Mohammed S. @foundmod]

@robinebers It is the opposite today, already consumed 50% of my weekly usage, only in two days!

[Mohammed S. @foundmod]

@said_azizow @wessyolo let us do it then tonight

[Mohammed S. @foundmod]

@ClimStefan No one is going to judge now keep making more videos you will find yourself gaining more confidence.

All support to you Clim

[Mohammed S. @foundmod]

@wessyolo We still didn't schedule, @said_azizow when we can host a space again?

[Mohammed S. @foundmod]

Before you make your last choice between Claude Code and Codex, read this.

I used to be a heavy Claude Code user, but recently i switched my workflows completely to Codex with GPT-5.5.

after 1 month, this is what i noticed.

Claude still wins frontend.

If i want a landing page, dashboard, hero section, or anything where the UI needs taste, Claude usually gets closer on the first try.

Better spacing.
Better layout.
Better visual decisions.
Less weird design.

Even after trying Opus 4.8, GPT-5.5 is still outperforming for my coding workflow.

Not in every task.

But for repo work, Codex feels more direct.

It reads the codebase.
Makes the change.
Shows the diff.
I review.
Then move on.

The biggest difference for me: Codex writes less redundant code.

Claude Code is very strong, but sometimes it adds too much.

Extra helper functions.
Extra components.
Extra abstractions.
Extra files i did not ask for.

The task gets solved, but the diff feels heavier than it should.

For content, i don’t see a huge difference.

Both can write decent drafts.

But Claude follows rules better. If i give it a strict writing style, banned words, examples, tone, structure, it usually respects the constraints better than GPT.

Claude Code has improved a lot too, especially with the recent code review updates.

So this is not “Claude can’t review diffs”.

It can.

But Codex still feels cleaner for my review loop.

The other thing i don’t like about Claude right now is how separated the workflow feels.

Claude Code.
Normal chat.
Cowork tab.

I know some people like having those separated.

I don’t.

It makes me think too much about where the work should live.

Is this a chat task?
Is this a code task?
Should this go to cowork?
Where did i leave the context?

With Codex, the loop feels simpler.

Task in.
Repo changes.
Diff review.
Next task.

I see a similar pattern in the community too.

People are not really picking one tool for everything anymore.

Claude for frontend, writing rules, and UI taste.
Codex for repo work, cleaner diffs, and code review loops.
Cursor for fast UI iteration and visual feedback.

So I don’t think the answer is “Claude is better” or “Codex is better”.

Right now, for my workflow, Codex + GPT-5.5 is winning.

If your goal is frontend, landing pages, designs, and polished UI, I would still choose Claude.

If your goal is working inside a repo, keeping code less bloated, and moving through diffs quickly, Codex has been better for me.

But this can change.

The tools are moving too fast.

Maybe next month Claude wins again.
Maybe Codex keeps pulling ahead.
Maybe we all end up using something else completely.

The final choice is not permanent.

It depends on the job in front of you.

[Mohammed S. @foundmod]

@thsottiaux Well, the real identity is built on Codex, so switching name will trigger the psycho to fell the quality drop.

So no

[Mohammed S. @foundmod]

@manuelsampedrop for real!

[Mohammed S. @foundmod]

Codex transcripts are the receipt layer!

If Copy as Markdown disappears, every agent run gets harder to audit, share, and turn into a build log.

I want boring export buttons more than another model picker.

[Simon Willison @simonw]

I'm really upset about this: OpenAI's Codex Desktop had a "Copy as Markdown" option for exporting full chat transcripts, but the feature vanished in an update a couple of days ago

Genuinely my single favorite feature of Codex compared to Claude Code
https://t.co/nk3yiPXHxL

[Mohammed S. @foundmod]

@CarlosBBuild Overdose of building

[Mohammed S. @foundmod]

@said_azizow that's cool, do you see high demand?