Muhammad Umar 🇺🇸

Founder CodeNinja, Helping America's Sovereign AI win in the Middle East and counterbalance China.

Connect :

[Muhammad Umar 🇺🇸 @Umargik]

The mission of CodeNinja is to help American sovereign AI win in the Middle East/emerging markets and act as a counterbalance to growing Chinese open source AI in the region.

@CodeNinjaInc @WhiteHouse

[U.S. Consulate General Lahore @USCGLahore]

American innovation is creating new opportunities for businesses around the world through strong commercial partnerships under President Trump’s leadership.

In Lahore, U.S. Chargé d’Affaires Natalie Baker met with Ahsan Jamil, Managing Partner of @CodeNinjaInc, a leading AI-powered software development company, to discuss how 🇺🇸 companies can expand by partnering with Pakistan's growing tech sector.

The conversation explored how American firms can leverage Pakistan's skilled tech talent through scalable delivery solutions, driving innovation, supporting U.S. business growth and strengthening commercial ties between our two countries. 🇺🇸🤝🇵🇰

[Muhammad Umar 🇺🇸 @Umargik]

@ArtificialAnlys This matters more outside Silicon Valley than in it. A lot of the world cannot afford the cloud bill, only the laptop. Tools like this decide whether the next Qwen download runs cheaper than the next American one.

[Muhammad Umar 🇺🇸 @Umargik]

@natolambert Agreed, and the harder problem is not banning it. America keeps under distributing what it will not ban. The Gulf and Asia are not choosing closed AI over open AI, they are choosing whichever open model actually shows up.

[Muhammad Umar 🇺🇸 @Umargik]

@superalesha The 8GB crowd is not just hobbyists. It is every developer in Nigeria, Pakistan and Indonesia who cannot afford a rack of H100s. Qwen already won that crowd, and nobody in Washington is tracking it.

[Muhammad Umar 🇺🇸 @Umargik]

Zhipu's new open model builds cyber exploits. Its safeguards fail up to 100 percent of the time, and anyone can download it.

That is the sovereignty question in one data point, via @AnthropicAI research. A model nobody can audit is not one anybody controls. https://t.co/3tgGNPKA5Z

[Ethan Mollick @emollick]

Leaving aside Anthropic's incentives for publishing this research, there is no doubt that open weights models will soon create the same security threats that closed source models have been demonstrating, except without guardrails. We are close. Probably good to plan accordingly.

[Muhammad Umar 🇺🇸 @Umargik]

Alibaba is expanding its AI data centres into the UAE. An Indonesian logistics firm already runs its finance workflow on Qwen.

America has Llama and Nemotron right there, free to download, and almost nobody here got sold one.

Not a model problem. A nobody showed up problem.

[Muhammad Umar 🇺🇸 @Umargik]

@CSIS Diversification is already happening in AI, not just defense.

Gulf states are running national models on Chinese weights while Commerce sits on 78 export proposals with nobody assigned to close them.

[Muhammad Umar 🇺🇸 @Umargik]

@DFCgov @DFC_CEO @semafor DFC closed the WIOCC deal because someone owned it end to end.

That is exactly what Commerce's AI export program is missing: 78 proposals ready, no account lead assigned to any of them.

[Muhammad Umar 🇺🇸 @Umargik]

@teortaxesTex @anton_d_leicht Dozens of entities shipping openly while Commerce sits on 78 proposals with no one assigned to close them.

That is the real scoreboard, not the benchmark chart.

[Muhammad Umar 🇺🇸 @Umargik]

Source: https://t.co/2uARTHSyNZ

[Muhammad Umar 🇺🇸 @Umargik]

Commerce got 78 company proposals in 90 days to export full American AI systems. Not one has an owner in Washington.

China shows up with one integrated offer. We hand buyers a stack of agencies and expect them to assemble it.

Fixable this quarter, if someone owns it.

[Muhammad Umar 🇺🇸 @Umargik]

@SemiAnalysis_ The depth of that breakdown is the tell. Nobody is writing infrastructure analysis this detailed about an American open model yet, because nobody outside China has shipped one that demands it.

[Muhammad Umar 🇺🇸 @Umargik]

@CNASdc @danielremler Export controls slow the chips. They do nothing to the download. Xi signed 29 countries to a rival open weight alliance this year, Pakistan and Indonesia included, and that front is wide open.

[Muhammad Umar 🇺🇸 @Umargik]

@allen_ai This is the honest version of open, data and recipe included, not just weights. The real gap left is distribution. Nobody is putting Olmo or Marin in front of buyers in Riyadh or Jakarta before the next Qwen drop gets there first.

[Muhammad Umar 🇺🇸 @Umargik]

Source: https://t.co/smKFyE9sKV

[Muhammad Umar 🇺🇸 @Umargik]

Xi launched a rival AI alliance. 29 countries joined, Pakistan and Indonesia included, Washington drew two dozen.

Chinese model usage was under 15% a year ago. Last week, over 54.

The fix is not a faster model, it is a stack open enough to read. We still have to build it.

[Muhammad Umar 🇺🇸 @Umargik]

@CarnegieChina Talent explains who builds the next model. It does not explain why Qwen and DeepSeek are already the default download from Jakarta to Lagos. That second race is distribution, and it gets a fraction of the attention the talent race does.

[Muhammad Umar 🇺🇸 @Umargik]

@G42ai @AiEverythingAD An ecosystem is the right word. The real test of it being sovereign is whether G42 and its customers can open any model in that system and say why it behaves the way it does, not just who is hosting it.

[Muhammad Umar 🇺🇸 @Umargik]

@VEONGroup Bangladesh is also one of the six countries in this week's MeetKai and Nvidia sovereign AI rollout through VEON. Capital plus a model trained inside the country by local engineers is the version of this that actually sticks.

[Muhammad Umar 🇺🇸 @Umargik]

Source: https://t.co/AuHZZDTCtz

[Muhammad Umar 🇺🇸 @Umargik]

An American company just shipped sovereign AI into Pakistan, Bangladesh, Kazakhstan, Uzbekistan, Ukraine and Brazil, national models trained in country, not a rented API.

Six countries in a week. That is the case study worth repeating everywhere Beijing already has a foothold.