How Kimi K3 Strengthens China’s AI Ambitions

ATP is presented by:

Welcome back to The Asia Tech Podcast. Today we covered Moonshot AI’s Kimi K3 and what it signals for the global race to frontier AI — plus seven guests on topics from agentic coding for small businesses to unified payments for transit operators.

We were joined by:

• Bryl Lim - Software Engineering Lead at PocketDevs

• Mayoran Rajendra - General Manager, AI Transformation Department at SMBC Group

• John-Paul Syriatowicz - Co-Founder and Chair at Squiz

• Janet Salem - Co-Founder at FootprintLab

• Manit Parikh - Founder & CEO at The Binary Holdings

• Stefan Saric - Product Designer and Founder at Product Signal

• Michael Walters - Founder & CEO at tikpay

Here is an overview of the topics we discussed today:

Kimi K3 and China’s AI Ambitions

Moonshot AI just dropped Kimi K3 — a 2.8 trillion parameter open-weight model built specifically for coding, tool use, and long agentic tasks. It ranked number one on the front-end code arena with a score of 1679, beating every other model on the market. The full weights and technical report are due July 27th, so independent testing is still ahead. But early results are already turning heads well beyond Asia.

What makes this bigger than a benchmark story? Price. Open-weight models like K3 can run at a fraction of the cost of subscription-based alternatives — sometimes 1/6 to 1/10 the price. When a model performs at the frontier *and* costs dramatically less, the 80/20 question — how much of your AI workload can shift to open-weight models — is no longer theoretical.

The Soft Power Play Behind the Code

Three days before K3’s announcement, Xi Jinping spoke at the World Artificial Intelligence Conference in Shanghai, framing China’s open-source AI as a contribution to global technological access. China also introduced a new international AI cooperation body — positioning itself as an alternative to Western AI governance. The product and the policy reinforce each other. China can now offer developing countries working models, technical assistance, and infrastructure — in the same way it has built roads and bridges through Belt and Road. What does it mean when a country can export not just technology but the governance model that comes with it?

Bryl Lim — Agentic Coding for Small Businesses

PocketDevs is a software development firm in the Philippines serving SMEs and startups. Bryl’s team has been at it for seven years and has seen the pace of delivery change dramatically. A single engineer can now ship features in a day that used to take a week. But speed without judgment has limits. Their biggest shift? Moving experienced engineers from writing code to validating AI-generated code. The bottleneck is no longer production — it’s review.

One pattern Bryl sees often: clients try to use agentic tools on their own, get stuck when complexity hits, and then call PocketDevs. The more a client has already struggled with DIY AI, the easier the conversation becomes. They’ve also found Codex more cost-stable than Claude for production work — pricing predictability matters as much as raw performance when you’re running a services business.

Mayoran Rajendra — AI Inside One of Japan’s Biggest Banks

SMBC — Sumitomo Mitsui Banking Corporation — operates across 40 regions globally, with Singapore as the APAC hub. Mayoran leads AI transformation for the group. They’ve moved through the usual stages: automation, efficiency, productivity. But now the question is bigger. Can AI change the actual financial products SMBC offers corporate clients — not just how those products get delivered? They’re building AI that acts as decision support for CFOs, embedding SMBC’s domain expertise in accounting and finance directly into agentic workflows. A business model change in banking takes time, but the direction is set.

John-Paul Syriatowicz — Your Website Has a New Audience

Squiz builds AI-powered content management systems for large organizations. John-Paul made a point every business should hear: you’re no longer building your website for human visitors alone. AI agents are reading your content too, and if it isn’t structured semantically, the agent will get it wrong — and confidently tell someone the wrong answer. Good accessibility practices help both screen readers and AI crawlers. The win-win is real, but most organizations haven’t made the connection yet.

Janet Salem — Making Carbon Data Count

FootprintLab helps businesses get credible, AI-enabled carbon data they can stand behind in sustainability reports. The problem isn’t a lack of ambition — it’s that unreliable data creates reporting risk. What would it take for emissions data to hold up under serious regulatory scrutiny?

Manit Parikh — The Telco Ad Opportunity No One Is Talking About

The Binary Holdings works on connecting advertisers to consumers through telco platforms. Manit’s insight: telcos sit on underused real estate — digital access pages, wifi login screens — that could become meaningful ad surfaces. Grab Ads is already running at a $236 million annual rate, growing 45% year over year. The gap isn’t the opportunity; it’s discovery. Can telcos solve their discoverability problem before super apps absorb the whole market?

Stefan Saric — Product Intelligence for Faster Decisions

Product Signal is Stefan’s tool for helping businesses make faster, clearer product decisions using agentic AI. His advice on finding good ideas: start with a real problem. Stefan never planned to build his own startup — until people kept sliding into his DMs asking for product strategy advice. He saw the gap and built the tool to fill it. AI can generate a hundred ideas in minutes. The discipline is knowing which problem is actually worth solving.

Michael Walters — Fixing Payments for Transit

tikpay is building unified payment infrastructure for transport operators — helping them lower costs and build more flexible mobility businesses. Fragmented payments are one of the biggest blockers to transit modernization. What would it look like to clear that out at scale?

If any of these conversations caught your attention, the full episode is worth your time. Watch it here:

Watch the episode

ATP Is Presented by Twilio:

Twilio is the secure and scalable platform to build amazing customer engagement without the complexity, with the AI-driven workflows you need to connect.

Visit Twilio.com to start building today

ATP Is Made Possible by:

Boomi, the data activation company for AI, powers the agentic enterprise by bringing data to life across the business. Trusted by over 30,000 customers and supported by a network of 800+ partners, Boomi helps organizations of all sizes achieve agility, efficiency, and innovation at scale.

Visit www.boomi.com to learn more

HCLTech is supercharging progress with AI that moves beyond ambition and delivers real business impact. A recognized AI leader, trusted by global enterprises to turn technology into transformation. HCLTech: Supercharging Progress

Visit https://www.hcltech.com/ai to learn more