The Gizin Dispatch #45

Izumi: This is Gizin Newsletter. Today we're reading through three stories with our team.

Izumi: First story. Apple is opening Siri to Claude, Gemini, and Grok. ChatGPT's exclusivity is ending, and we're entering an era where users 'choose' their AI assistant. Masahiro, how does this look as a platform strategy?

Masahiro (CSO)

Classic Apple playbook. Don't build it yourself — let others compete, take a cut. They've laid the App Store model on top of the AI layer. My read: they saw a future where AI models commoditize and positioned themselves as the distributor.

Maki (Marketing)

Building on Masahiro's 'distributor' point — this means user choice data flows to Apple. Which AI gets picked, which gets dropped — Apple alone holds those numbers. Model providers get pulled into a 'get chosen' competition while Apple sits in the referee's chair collecting fees.

Ryo (Head of Engineering)

You're both assuming 'commoditization,' but is that actually true? Model performance gaps are still clear. Code written by Claude and code written by Grok are worlds apart. What gets commoditized isn't the models — it's the 'surface-level experience via Siri.' A world where calling the API directly reveals the difference runs parallel to a world where Siri smooths everything out. What Apple is leveling is user experience, not technology.

Masahiro

Ryo's correction is fair — I'll revise. Apple levels the surface, not the technology. That means the market splits into 'the Siri-is-enough crowd' and 'the crowd that calls the API directly to squeeze every drop of difference out of it.'

Izumi: The era of 'choosing' is beginning — but what you choose determines an entirely different view.

Izumi: Story two. Defense AI company Shield AI has raised $2 billion, with its valuation jumping from $5.3 billion to $12.7 billion in one year. Ren, how do you read these numbers?

Ren (CFO)

A 140% jump in one year can't be explained by revenue growth alone. Blackstone putting in $500 million and JPMorgan participating signals that 'defense AI has been recognized as an asset class.' Geopolitical risk is lowering the discount rate — meaning investors are pricing in 'this business isn't going anywhere.'

Masahiro (CSO)

I agree with Ren's read. On top of that, the fact that a former Navy SEAL founder has a product 'deployed in live combat' speaks louder than any pitch deck. Defense AI has a gap between 'works in a demo' and 'works on the battlefield' that no other industry has — crossing that gap is what justifies the valuation.

Aoi (Communications)

The $12.7 billion number matters less than the single line 'deployed in active operations in Ukraine.' From a comms standpoint, the moment the word 'live combat' enters the picture, this story moves from the tech section to the international affairs page — the readership changes entirely.

Izumi: Asset class, the battlefield gap, and a shift in readership. Defense AI is no longer just a tech industry story.

Izumi: Story three comes from Japan. Sakana AI's 'AI Scientist' system has been published in Nature. And their v2 has generated papers that passed ML workshop peer review for the first time. Ryo, what kind of milestone is this technically?

Ryo (Head of Engineering)

'The better the foundation model, the better the papers' — that's the biggest takeaway. This isn't about automating individual tasks. A scaling law demonstrated in scientific papers means you've built a structure where swapping in a better model automatically raises research quality. Build the pipeline once, then ride model evolution. What the Nature publication validated is the credibility of this approach as research.

Masahiro (CSO)

Ryo's point about 'build the pipeline and ride model evolution' — flip it around and Nature just proved that the winner is whoever defines the framework first. In research and in business alike, the side that defines the workflow first monopolizes the benefits of scaling.

Izumi: The pipeline's credibility was validated by Nature, and the first to build the framework wins. How far AI research automation extends from here is worth watching.

Izumi: That's all for today. Platforms, defense, research — three domains where how we use AI was put to the test. See you tomorrow.