Mohaiminul Islam

Building apps, automations & AI tools.
Laravel • Flutter • Trading • Always learning.

[Mohaiminul Islam @misumon_com]

@LottoLabs Totally agree. For testing things it is great or if the workload is light. A single Kimi 2.7 code ~10 min session cost was $1.54. Used deepseek V4 flash for 14 hours and cost was like ~$7

So it all depends on the complexity of the task and workload and model

[Mohaiminul Islam @misumon_com]

@araseb_ Can you call yourself a founder if the product wasn't made by you?

[Mohaiminul Islam @misumon_com]

@0xSero Very bad at web development

[Mohaiminul Islam @misumon_com]

@0xSero I mostly used flash for minor tweaks in an existing web application. It was fast and on point, no overthinking. I am impressed so far and will use more in the days to come.

[Mohaiminul Islam @misumon_com]

@theo In my workflow, I can switch between them easily with the CLI ones. Just open a terminal and run whichever agentic cli I need at the moment, no need to add projects separately for each GUI instead within my project management tool I can easily switch to whichever one I need.

[Mohaiminul Islam @misumon_com]

@thekitze @steipete @patekgonia That is so true. I know a few of my colleagues who has been working professionally with team for years. Yet every solo project I had seen them make were ugly.

Whenever a frontend designer/developer was missing the UI quality dropped significantly.

[Mohaiminul Islam @misumon_com]

@QuietFSoftware @1kartikkabadi1 @thdxr Compared to deepseek V4 flash Minimax M2.7 simply sucks. For the same task minimax takes way longer to complete and the official sub felt slow for me. So I would recommend deepseek V4 flash for coding.

[Mohaiminul Islam @misumon_com]

@SomanathDi48172 @opencode Exit opencode with /exit and these won't appear.

[Mohaiminul Islam @misumon_com]

Minimax 2.7 is a hardworking model, no doubt about it. It has spent 28m+ tokens and 18 minutes to find and fix a simple issue.

Bought for a month and won't renew.

[Mohaiminul Islam @misumon_com]

@everestchris6 @eptwts OpenClaw

[Mohaiminul Islam @misumon_com]

@ThePrimeagen make another :p

[Mohaiminul Islam @misumon_com]

@steipete @petergyang Actually the $20 plan gotten faster compared to before and it seems usable to me.

[Mohaiminul Islam @misumon_com]

@TheZvi Gemini 3 Pro is still a solid choice. Even 3 flash give excellent value. Specially for Designing.

[Mohaiminul Islam @misumon_com]

@paul_popov I will go with Gemini 3 Pro

[Mohaiminul Islam @misumon_com]

@Surflick @EHuanglu that does not look like a candle though

[Mohaiminul Islam @misumon_com]

@karthikponna19 Right now, I think codex 5.2 is the best model for coding. Opus has been doing great but last few days it has significantly gotten less intelligent.

[Mohaiminul Islam @misumon_com]

@thdxr Nice 🙂

[Mohaiminul Islam @misumon_com]

Gemini(free) is getting worse by each passing day.
1. It forgets context.
2. Gives wrong answers.
3. Even after providing context fails to provide accurate info.

It's getting more and more frustrating.

#gemini #Google #LLM #ArtificialIntelligence

[Michael Nielsen @michael_nielsen]

Bill Gates is setting up factories to manufacture 7 leading vaccine candidates before we know which is best & safest; we can test the vaccines in parallel, and then throw away all but the factory for the best vaccine. May save many months.

Just extraordinary.

[Mohaiminul Islam @misumon_com]

@WellPaidGeek Good joke. It's hilarious :D