Good morning. Here is what matters in AI today, and how to put it to work.
We see AI agents earning trust in security ops while consumer AI struggles to hold attention, and data center politics drift from evidence.
~3 min read · last 12 hours
In today's issue
01
Figma deploys AI agents to automate security work
02
Apple's revamped Siri: early excitement, rapid forgetting
03
Tech leaders lean on China fears to sell data center expansion
Main story
Figma deploys AI agents to automate security work
Figma's engineering team published how they built AI agents to handle security tasks, offering a concrete, documented example of agentic AI earning its place in a production workflow.
Why it matters: This is one of the cleaner public blueprints for AI-augmented security operations: if your team is scoping agent use cases, Figma's approach is worth studying as a credibility benchmark.
What to watch next: Watch for other engineering teams to publish similar agent playbooks: the more documented production cases accumulate, the faster organisations can move from pilot to justified deployment.
We see a clear split today between AI delivering real operational value inside engineering teams and AI struggling to hold end-user attention, and both signals matter for how you prioritise your roadmap.
Today's items collectively point to a maturing AI landscape where the gap between production value and consumer promise is widening. Agentic AI is finding its footing in internal, well-scoped engineering problems like security, where outcomes are measurable and the human-in-the-loop is natural. Consumer-facing AI, by contrast, is fighting a retention battle that polish alone will not win. Meanwhile, the infrastructure narrative driving AI investment is increasingly political, and teams relying on that tailwind should watch for the moment the evidence gap catches up with the rhetoric.
All the best, the KYFEX team
Quick hits
AI in production: security agents and fading consumer hype
Apple's revamped Siri: early excitement, rapid forgetting
A hands-on account finds that the beta of Apple's AI-powered Siri generated initial enthusiasm but quickly faded from daily habit, raising questions about whether the full release can sustain engagement.
Why it matters: Consumer AI features live or die on stickiness, not launch buzz: product teams building AI-powered assistants should treat retention, not first-impression delight, as the real success metric.
Data center politics: China as a convenient, evidence-free villain
The framing of China as the reason America needs more data centers is gaining political traction even as the factual basis for it remains thin, and that gap has real consequences for infrastructure policy and public trust.
Tech leaders lean on China fears to sell data center expansion
Polls show most Americans oppose new data centers in their communities, and analysis finds that invoking China as a national-security justification is a rhetorical strategy largely unsupported by evidence.
Why it matters: If your organisation is making infrastructure or policy arguments, grounding them in verifiable facts rather than geopolitical fear is both more durable and less exposed to credibility risk when the evidence is scrutinised.
Learn about AI technologies and how to use them. Examine AI's societal impact, and learn to navigate this technological shift.
DeepLearning.AI · Free · 1 hour
Put it to work
Try this today
Scope an AI security agent for your team
You are a senior security engineer. I will describe our current manual security review process step by step. For each step, tell me: (1) whether an AI agent could automate it reliably today, (2) what the key failure mode or risk is, and (3) what human oversight checkpoint I should keep. Here is our process: [paste your process here]. Be direct and flag any steps where you would not recommend automation yet.
Why it helps: With Figma's agent blueprint now public, this prompt helps your team do a fast, honest readiness assessment before committing engineering time to a security agent build.
Before you ship it
The risk
Security agents operate on sensitive data and access controls, and an overly broad agent scope can expose vulnerabilities or make automated decisions that should require human sign-off.
Do this
Define explicit, narrow permissions for each agent action at design time, and require a human approval step for any action that modifies access controls or escalates privileges.
Ready to ship AI, not just read about it?
KYFEX designs and builds production AI for teams that need it working, not just demoed. Tell us what you're working on and we'll bring the engineering.
