Analyse Podcast

Fresh out of the studio, Ang Li, CEO and co-founder of Simular and previously a research scientist at Google DeepMind, joins us to explore why computer use is the last mile to AGI. Ang traces his path from studying catastrophic forgetting and continual learning at DeepMind to founding a company on a single conviction: models break in production because the data distribution never stops moving, and the only place to close that loop is the real world. He explains why the chatbot metaphor misleads enterprise buyers, why the right unit of measurement is tokens per task rather than price per token, and how the power law of practice should make an agent cheaper every time it repeats the same work. He separates capability from reliability through pass@k and pass^k, argues that deterministic work belongs in code rather than in a model, and makes the case that frontier labs exist to sell tokens while Simular exists to remove them. Last but not least, Ang shares his test for AGI — one you feel rather than see — and what it will take to bring autonomous work down to the cost of water.
"So when I go to the gym, [doing] the same workout everyday, as I do this more and more, more reps, more sets, I don't even need to think. It's called muscle memory. I just do the same in a standard way, my consumption or my tokens in my brain dramatically drops. And there's a term describing this behavior in human cognition called 'the Power Law of Practice'. So that's the whole idea. Do we see power law practice in AI agents? We don't have that yet."Profile: Ang Li, CEO and co-founder of Simular AI
X: https://x.com/angli_ai
LinkedIn: https://www.linkedin.com/in/angli-ai/
Simular AI: https://simular.ai
Episode Highlights:
[00:00] Quote of the Day by Ang Li from Simular
[01:09] Why AGI arrives through the mouse and keyboard
[02:27] Legacy systems with no APIs block automation
[03:45] Computer use is the last mile
[04:05] One device, a hundred agents in the cloud
[05:51] The gap between research and production
[07:45] The day Covid became a dominant search keyword
[08:13] Continual learning over evolving distributions
[09:21] Catastrophic forgetting remains unsolved
[10:11] Why the learning loop requires a product
[11:04] Why AGI must be built commercially, not academically
[12:23] Chat versus work: two different problems
[13:40] Every company needs a repeatable playbook
[15:10] Tokens per task beats price per token
[16:13] Same task 100 times, 100 times the tokens
[16:34] The gym analogy and the power law of practice
[18:18] Where the scaling frame breaks for computer use
[20:46] Why AGI will not be a single model
[21:43] The 72.6% peak versus repeated-run reliability
[22:32] Hiring analogy: the interview versus daily work
[25:22] Replaying trajectories to drive token cost down
[26:17] The idea almost nobody in the industry discusses
[29:01] Democratising autonomous computers to the cost of water
[30:26] SMBs, not enterprises, are the real market
[32:59] Why healthcare and insurance came inbound
[34:06] Launching Sai on Windows virtual machines
[34:38] Simulang and the unity of opposites
[35:52] Frontier labs sell tokens; Simular removes them
[38:01] The hidden cost of self-hosting open models
[39:23] Why CPU and memory matter more than GPU
[41:10] You only have two hands: the physical constraint
[42:48] The AGI test you feel rather than see
[44:19] Why humanoids take longer than software
[45:23] The motivation: doing your work from the forest
[47:03] What Great Look Likes for Simular AI
[48:34] Closing
Podcast Information: Bernard Leong hosts and produces the show. The proper credits for the intro and end music are "Energetic Sports Drive." G. Thomas Craig mixed and edited the episode in both video and audio format.

Fresh out of the studio, recorded in Singapore, Steve Clayton, Senior Vice President and Chief Communications Officer at Cisco, joins us to discuss what it takes for AI to scale inside enterprises globally. Steve argues that narrative belongs at the top of the AI pyramid, not the bottom — in an era when models generate more material than any organisation can absorb, the story a company can credibly tell becomes the scarce asset, not the technology itself. He explains why the truth lies in the field rather than at global headquarters, how regional dynamics reshape where a story lands, and walks through Cisco's advances in silicon and quantum computing networking. Showing a real customer deployment, he argues, is what earns permission to talk about security and trust at all. Last but not least, he shares what great look like for Cisco to build their story in the age of AI.
"Take your phone and switch it into airplane mode and see how much you can do on your phone. Very little. You can use the calculator, but you can't use an AI model. You can't surf the web, you can't send messages. And people get excited about GPUs and data centers and models and applications and devices. But sitting at the center of all of that is the network. And that's what Cisco is all about. So that's our job, is to help people understand the value of the network and to realize that AI, as exciting as it is, it simply doesn't happen without the network." - Steve Clayton, SVP & Chief Communications Officer, CiscoProfile: Steve Clayton, SVP & Chief Communications Officer, Cisco
LinkedIn: https://www.linkedin.com/in/stevecla/
Friday Note Newsletter: https://www.linkedin.com/newsletters/the-friday-thing-7412856034161229824/
Episode Highlights:
[00:00] Quote of the Day by Steve Clayton from Cisco
[00:30] Introduction: Steve Clayton from Cisco
[02:13] Why storytelling began around the campfire
[02:55] Telling stories about impact, not technology
[03:43] Why the AI debate over-indexes on models
[04:51] The airplane mode test for AI dependence
[06:17] Steve inverts the pyramid, narrative on top
[07:04] Communicating without leading with the technology
[07:49] Breaking through by doing the opposite
[08:37] Why "content" has become a bad word
[09:59] Audience-first: meeting policymakers where they are
[11:10] The trap of filling the pipeline
[11:34] Attention, not creativity, is the scarce resource
[12:15] Curation and judgment against AI slop
[12:51] We are not entitled to people's attention
[13:26] Silicon One, the Cisco business nobody expects
[14:26] Cisco's universal quantum switch at room temperature
[14:57] Balancing ambition against credibility
[16:08] Show the possible, do not just describe it
[16:35] Earning permission to talk security and trust
[17:18] The truth lies in the field
[18:16] McLaren F1 and Networking Academy stories
[19:12] The question Steve wishes people would ask
[20:56] The Friday Thing: human first draft, AI editor
[21:26] Humans create the novel ideas, AI supports
[22:07] What Cisco should mean to business and to family
[23:21] Closing
Podcast Information: Bernard Leong hosts and produces the show. The proper credits for the intro and end music are "Energetic Sports Drive." G. Thomas Craig mixed and edited the episode in both video and audio format.

Fresh out of the studio at the Dataiku Summit Singapore 2026, Nur Hafiza Mutalif, Assistant Head of International Affairs at the Singapore Red Cross, joins us to explore why the humanitarian sector distrusts AI, and what her team built once they worked through that scepticism. Coming to this from international law rather than data science, Nur Hafiza explains why the sector's caution is a feature rather than a lag: humanitarian work rests on doing no harm and on neutrality, and any solution must clear that bar before it clears a business case. Dataiku cleared that bar. Its pro bono experts taught the team how the models worked and let them set the boundaries — which is what earned their confidence. Hafiza walks through the disaster surveillance pipeline that freed four people from daily collation, and the Thailand leptospirosis model forecasting outbreaks from case and climate data. She argues AI should never be justified on efficiency, that data efficiency itself carries risk, and that acting on a forecast demands low-regret choices. Closing the conversation, she shares why capability outlasts any model.
"We cannot see efficiency, too, I think what I should see, as a humanitarian organization, is datasets and the different cases around the world be compressed into something that simplifies value and help us better serve communities." - Nur Hafiza MutalifProfile, Nur Hafiza Mutalif, Assistant Head of International Affairs, Singapore Red Cross
LinkedIn: https://www.linkedin.com/in/nurhafizaab/
Episode Highlights:
[00:00] Quote of the Day by Nur Hafiza Mutalif from Singapore Red Cross
[00:55] Three angles: distrust, what they built, where the line sits
[01:45] From international law to leading an AI project
[02:53] What made a cautious team comfortable with AI
[03:24] Pro bono experts who did not impose solutions
[04:04] Attention is a luxury for humanitarian organisations
[05:13] "We curate, we don't create" — countering compassion fatigue
[05:59] Do no harm, neutrality, and innovating with care
[07:06] Why AI must not be treated as an efficiency tool
[07:55] Disaster surveillance: four people, days reclaimed quarterly
[09:09] Does AI only automate the boring work?
[10:24] The Thailand leptospirosis outbreak prediction model
[11:21] Getting humanitarians comfortable acting on a prediction
[12:47] Where AI changes the disaster cycle most
[13:44] What outsiders get wrong about data efficiency
[14:38] The line AI must not cross at the Red Cross
[15:40] Why internal capability beats the models themselves
[16:33] Where an organisation with no data team starts
[17:19] The question people ask, and the one they should
[18:04] AI as a cross-cutting theme over five years
[19:01] Closing

Podcast Information: Bernard Leong hosts and produces the show. The proper credits for the intro and end music are "Energetic Sports Drive." G. Thomas Craig mixed and edited the episode in both video and audio format.

Fresh out of the studio, Sophie Dionnet, Senior Vice President of Product and Business Solutions at Dataiku, joins us at the Dataiku Summit in Singapore to discuss what turns enterprise AI investment into measurable value. She lays out the three ingredients Dataiku builds around — the right people, orchestration across technologies, and supporting controls — and makes the case that governance is a scaling mechanism rather than a brake. She points to Roche, where a patent lawyer encoded his own professional expertise into a working system of agents, discusses Dataiku's answer to agent sprawl with agent management launching in October, and closes on strong momentum across banking and the public sector in Asia Pacific."A lot of the changes that organizations need to do today actually don't require the latest model. That's not really the problem. It's about doing the hard thing, the change, the things that we talked about. It's easier to be excited by the new toy than by trying to use it. And so yes, I think this is why there is a bit of a gold rush of trying to figure out where is it going to end. We don't know." - Sophie DionnetProfile: Sophie Dionnet, Senior Vice President of Product and Business Solutions at Dataiku
LinkedIn: https://www.linkedin.com/in/sophie-dionnet-a176894/
Episode Highlights
[00:00] Quote of the Day by Sophie Dionnet from Dataiku
[01:00] Three angles: domain knowledge, orchestration, governance
[01:59] What has not changed: data still decides everything
[02:31] Data consciousness accelerated over the past twelve months
[03:05] The LLM explosion and the raw-power question
[03:51] Why Sophie pushed governance before the market asked
[05:30] What Dataiku is, and where the name comes from
[06:15] Three ingredients: people, orchestration, controls
[07:22] Roche: a patent lawyer builds his own agents
[08:51] Change management, not technology, is the gap
[09:41] Decision takes an hour, implementation takes two years
[09:58] Why domain knowledge beats model performance
[11:30] Most changes do not require the latest models
[11:58] The scaling belief the industry gets wrong
[12:50] Centralisation risk and the rise of shadow AI
[14:02] Where leaders still quietly choose to do nothing
[15:22] Vibe coding, conflicting outputs, and lost consensus
[16:43] The GDPR lesson on ex-post compliance cost
[18:42] Why the agent question starts at the board
[19:38] Agents are simply a new kind of API
[20:28] Is agent sprawl technology or organisational design
[21:28] What separates AI scalers from pilot purgatory
[22:58] The bear case: foundation labs absorb the middle
[23:45] Why every leading technology becomes self-centred
[24:57] Vibe coding your own Salesforce, and why not
[25:23] The pet store analogy for build versus buy
[26:22] Systems of record and the real switching cost
[28:30] Dataiku in Asia Pacific over the next three years
[29:59] Closing

Podcast Information: Bernard Leong hosts and produces the show. The proper credits for the intro and end music are "Energetic Sports Drive." G. Thomas Craig mixed and edited the episode in both video and audio format.

Fresh out of the studio, Dr. Rafael Frankel, Head of Public Policy for Asia Pacific at Meta, joins us to unpack how Meta is stepping up on the fight against scam syndicates in Southeast Asia along with a coalition of partners. He explains why today's syndicates operate with state-like capabilities and forced labor at scale and how AI has become both the scammers' sword and Meta's shield, powering the removal of hundred million scam accounts in 2025. Frankel walks through the disruption-sprint model that took down the million accounts in a single week alongside the FBI, Microsoft and Starlink, and argues that beating scams demands a coalition matched to the criminal supply chain. Closing out, he reminds us that crime is never zero — but through sustained, collective pressure, it can be cut by orders of magnitude.
"I think that people underestimate the criminals that are the drivers of the scams. Mostly what we're talking about at this point is extremely well-financed, highly sophisticated transnational criminal networks that have turnover in the billions, tens of billions, and have the finances, the technological know-how, the sophistication to really be trying to scam people 24/7, 365 days a year. That's a tough adversary to be up against." - Dr Rafael FrankelProfile: Profile: Dr Rafael Frankel, Head of Asia-Pacific Public Policy at Meta
LinkedIn: https://www.linkedin.com/in/rafael-frankel/
Episode Highlights:
[00:00] Quote of the Day by Rafael Frankel from Meta
[02:38] From foreign correspondent to tech policy
[03:39] Career lessons: gut instinct and taking risks
[04:33] Why the Meta APAC policy role, and why now
[07:26] The one thing people misunderstand about scams
[09:13] What changed: industrial scale scam centers
[11:35] AI as Meta's shield, deployed at scale
[13:07] Genesis: the Bangkok 2024 convening
[14:14] First disruption: 59,000 accounts, six warrants
[15:45] June haul: 1.5M accounts, Microsoft, Starlink, 63 arrests
[16:52] Matching the coalition to the criminal supply chain
[18:52] Building trust: results create the feedback loop
[20:05] The Global Signal Exchange and the FIRE program
[22:50] Counter-intuitive lesson: Meta doesn't see everything
[24:13] Asia Pacific as front line of the solution
[26:20] Pulling on the same rope across governments
[27:30] POGO centers and pinpointed police action
[29:02] Is sanctioning a kingpin just theatre?
[30:52] Crime is never zero, but reducible by orders of magnitude
[31:02] The question Rafael wishes people asked: how can I help?
[32:52] What success looks like three years out
[35:08] Closing
Podcast Information: Bernard Leong hosts and produces the show. The proper credits for the intro and end music are "Energetic Sports Drive." G. Thomas Craig mixed and edited the episode in both video and audio format.