Katie Thomas

@FARLabsAI How much does this really depend on better inference efficiency?

As software value moves toward agents, demand moves toward inference. The winners will be the networks that make AI faster, cheaper, and usable in real time.

@Dizzaract On-device models could change a lot if they become practical at scale.

AI in game development is still mostly theater. 36% of studios claim adoption, but 52% report workflow damage. Artists and designers resist for obvious reasons, most implementations are surface-level bolts-ons. The real work is happening at the architecture level. On-device models. AI-native design from day one. 2026 is when you'll see the difference between studios that added AI and studios that were built on it. We're building the latter.

@FARLabsAI How much of this gets solved through hardware vs better model efficiency?

@FARLabsAI people are starting to see it

@FARLabsAI Looks 🔥

DLSS 5 is AMAZING. The future of AI inference has never been clearer.

@shzhv13 Technology keeps removing friction

AI replacing middlemen is inevitable... Most service industries are just information arbitrage with 30 years of regulatory capture in my opinion

@FARLabsAI Gaming definitely trains reaction time

Centralized infrastructure limits what's possible. Every AI app runs through the same 3-5 cloud providers. When everyone builds on identical infrastructure, innovation converges. ❌ Same latency constraints ❌ Same scaling patterns ❌ Same cost structures ❌ Same geographic limitations ❌ Same architectural decisions ❌ Same bottlenecks. Distributed compute opens new design space.

@shzhv13 45. Definitely coming back to this one.

This took me 10 years to learn: Most companies optimize for the wrong timeline. You don't just need quarterly wins. You need: - revenue that compounds - infrastructure that lasts - IP you actually own - systems that get stronger with time I've watched companies chase vanity metrics. They hit their slide deck numbers. Raise another round. Then the market shifts and they're gone. Meanwhile, the ones building for decades own the entire stack. They're not worried about impressing the next VC. They're building systems that print money in 2035. Revenue matters. But sustainable revenue matters more. Stop optimizing for the next pitch deck. Start building something that outlasts you.

@shzhv13 Interesting post, Ilman. Any thoughts on whether this puts pressure on other chip makers like AMD to create similar software ecosystems?

🤯Nvidia is giving away a billion dollar AI product for free. Here's the play: Free platform → requires massive compute → Nvidia sells the compute. Give away the razor. Sell the blades. The AI agent market hit $10.9B in 2026. 40% of enterprise apps will embed agents by year-end (up from 5% last year). 85% of companies already use AI agents. 93% plan full deployment within 2 years. Nvidia isn't competing on products anymore. They're controlling the infrastructure everyone else has to rent. Training chips → inference chips → deployment platforms. Own the bottleneck. Make everything else free. Classic vertical integration disguised as open source. The pattern is simple: when you control multiple layers of the stack, each one makes the others more valuable. The companies that win won't have the best product. They'll own the stack everyone else pays to use. NemoClaw likely launches next week at GTC. Watch what happens.

@shzhv13 Interesting thesis but the latency constraints on distributed consumer-device inference are non-trivial. How are you handling model partitioning across heterogeneous hardware?

Simple observation. Anthropic users consume $5K monthly in API value. Charge them $200. Mathematics is clear. This doesn't scale. Yes, $5K is retail. Real compute cost lower. But centralized inference creates structural problem that capital cannot solve indefinitely. 2026: inference will be 85% of enterprise AI spend. Infrastructure determines survival. Not marketing. Not features. Infrastructure. I built Bitcoin mining facilities in 2017. Pattern was identical every time. Control infrastructure cost = survive. Don't control it = die when market corrects. Physics doesn't care about your pitch deck. AI follows same laws. Centralized compute is expensive. Will always be expensive. VC money delays the problem, doesn't eliminate it. Solution is obvious: distributed inference across consumer devices. Cost per token drops Margins scale properly Economics work without constant fundraising We build this from Abu Dhabi. Infrastructure layer decides everything. This is not opinion. This is physics. Markets will prove this correct or incorrect. I believe correct. Time will show.

@gamedofficial_ Happy International Women’s Day to everyone in the industry 🙌

Happy International Women's Day from GAMED! 🩷 Tell us who some of your favorite women in gaming are?

@Dizzaract Happy International Women’s Day ❤️

The future of AI and gaming is being written right now and women are helping lead it. Developers, founders, artists, researchers, community leaders. New ideas, new worlds, new technologies. Today we celebrate the women pushing innovation forward. Happy International Women’s Day!

@shzhv13 Interesting how fast public sentiment can swing after a decision like that.

Anthropic refused $200M Pentagon deal Thursday. OpenAI took it Friday. 48 hours later: OpenAI: → ChatGPT uninstalls up 295% → 1-star reviews surged 775% → Downloads fell 13% Anthropic: → Claude hit #1 App Store → 149K daily downloads → Record-breaking signups all week Anthropic got banned from federal contracts. The public made them #1 anyway. Lesson here is stay ethical and hold to your principles. You will never lose out.

@FARLabsAI Reliability is everything if people are running real workloads.

Every GPU provider on the FAR AI network will only be rewarded for work that's actually verified. Here's how that works: When your GPU completes a workload, an orchestrator checks the result before it counts toward your earnings. This is called proof of compute the network's way of confirming that real work was done, correctly, before any earnings are assigned. It keeps the network reliable for the people running AI on it, and fair for the GPU providers powering it.

@shzhv13 Totally. The winners will be the teams with clean data + fast iteration loops.

Multi-agent AI systems are crossing the economic viability threshold $200K salary vs $2K/month compute. The transition isn't theoretical anymore. It's happening in production environments right now. The question is deployment speed and integration quality

@Dizzaract @CapcomUSA_ The compounding IP point is the real takeaway here.

Resident Evil Requiem has achieved $350 million in revenue from 5 million units sold in just five days. Congratulations to @CapcomUSA_ and the entire development team on this remarkable launch. This performance demonstrates the enduring power of well-crafted intellectual property and the value of long-term investment in world-building. After 28 years, the Resident Evil franchise continues to set benchmarks for the gaming industry. What stands out is the compounding nature of IP development. Each release builds on decades of narrative depth, community engagement, and technical innovation. This isn't just a successful game launch - it's the result of consistent investment in a universe that spans multiple platforms, media formats, and generations of players. The gaming industry is at an inflection point. While short-term models dominate headlines, launches like this remind us that the most valuable assets in gaming are built over decades, not quarters. At Dizzaract, we're building with the same long-term perspective. Original IP that creates new cultural aesthetics. Infrastructure that scales with the ecosystem. Systems designed to compound value across multiple product cycles. Achievements like Resident Evil Requiem reinforce why we're committed to patient, strategic development in both gaming and AI infrastructure. The future of gaming belongs to those who build for decades.

@Dizzaract Hoping for a quick return to normal for everyone

In these challenging days, we are deeply grateful for the UAE’s swift and professional activation of its National Emergency, Crisis and Disaster Management System. The tireless dedication of our military, civil authorities, and frontline teams continues to keep residents, businesses, and innovators safe and supported across the country. We are proud of the genuine unity we see every day -Emiratis and our international colleagues standing together, checking in, and moving forward with calm determination. This shared resilience strengthens the collaborative spirit that powers our work in AI and gaming right here in Abu Dhabi. Thank you to everyone protecting lives and sustaining progress. The UAE’s preparedness and community strength inspire us all. From all of us at Dizzaract 🇦🇪

@shzhv13 Specialization really is where things are heading

General models try to do everything, so they're average at most things. A model built for one person's actual workflow doesn't waste resources on features they'll never touch. Specialization is becoming the focus.

@shzhv13 Strong message. Unity always matters most in moments like this.

From the heart of Abu Dhabi, I send my solidarity to all across the UAE. This country is built on unity and mutual respect. In times like this, those values guide us.

@shzhv13 $519M in UAE alone is serious momentum

The AI tsunami is coming. And I'm tracking a critical moment unfolding in MENA. Dario Amodei just issued a warning the world is ignoring: → Superhuman AI by 2027 → 18-month window before total transformation → Complete automation of cognitive labor He's not winning because he has the best tech. He's winning the "trust war" - because he's brutally honest about what's coming. Here's what I'm seeing in our region: MENA AI funding hit $858M in 2025. UAE captured $519M (up 267% YoY). AI is projected to contribute $320 billion to MENA's economy by 2030 - roughly 11% of regional GDP. UAE is building the world's largest AI campus outside the US - 26 sq km in Abu Dhabi with 5GW capacity. The capital is flowing. The ambition is real. But here's the challenge ahead: Despite massive infrastructure investments, governance frameworks are still maturing across the region. Nearly half of organizations cite talent shortages and insufficient technological capabilities as barriers to scaling. Complex AI technologies require specialized understanding. Many AI projects remain in testing phases. We're building the infrastructure. Now we need to match it with readiness. The questions that need answering: When AI automates 50% of entry-level jobs, are our systems optimized for scale? When inference costs become prohibitive, who's building the distributed infrastructure? When the software industry transforms overnight, what does that mean for MENA? Logistics? Finance? Government services? I've spent the last 10 years building to address these challenges. Most regional players are celebrating AI's potential. I'm focused on: "Here's how MENA can lead through this transformation." At Dizzaract and FAR Labs, we're not building AI products. We're building infrastructure for regional readiness. FAR AI is our answer to expensive, centralized inference: → Distributed compute network across consumer devices → Cheaper, faster, and scalable AI inference → Infrastructure that serves any AI application globally - without dependency on foreign data centers We're building from the ground up. From Abu Dhabi. For the region. My team of 80+ engineers, researchers, designers, and operators are building the technical systems the region will need: → AI-native architecture designed for sovereignty → Full-cycle control from IP to infrastructure → Execution in weeks, not quarters This is what we're building for. In the coming months, when Ministers and CEOs ask: "How do we scale AI without losing sovereignty? Without becoming dependent on foreign infrastructure? Without sacrificing our values?" They'll need partners who aren't just talking about readiness, but are building the systems. Partners who understand both the technology AND the territory. Partners who are honest about the challenges, while everyone else sells only optimism. Amodei admitted "no one elected me to make these decisions." That vacuum exists in every region. And I'm committed to being part of the solution here. The future of AI in MENA won't be written in California. It's being written here. Right now. By people who saw this coming, and are building the infrastructure to meet it. The clock is ticking.