Ryan Loots

Connecting dots in tech, AI, and business. Writing what the feeds miss.

[Ryan Loots @_rloots]

https://t.co/gaAhhP4eqc

[Ryan Loots @_rloots]

https://t.co/0Wrv49Gn1X

[Ryan Loots @_rloots]

https://t.co/ZneryT4G7M

[Ryan Loots @_rloots]

https://t.co/PGpTC2N1b2

[Ryan Loots @_rloots]

https://t.co/S3FPog11AB

[Ryan Loots @_rloots]

https://t.co/aYg0mUuyw5

[Ryan Loots @_rloots]

The 60% / 0-20% number is the real story in that report.

Sixty percent of dev work touches an agent. But fully delegated tasks: zero to twenty.

That's not autopilot. That's supervised acceleration. The teams who treat the gap as engineering work, not a model problem, are the ones quietly pulling ahead.

Wrote up the full pattern: https://t.co/o6E7Kx4hRP

[Ryan Loots @_rloots]

https://t.co/0ddkar4HXg

[Ryan Loots @_rloots]

https://t.co/0ddkar4HXg

[Ryan Loots @_rloots]

@deftunix @HikmaAi_ If your agent can't obey these, it's not an agent. it's a chatbot with anxiety.

[Ryan Loots @_rloots]

https://t.co/9vNg31I1pz

[Ryan Loots @_rloots]

Apple's Guideline 2.5.2 isn't anti-AI. It's the first platform-level governance policy for AI-generated code.

"An app must not change its functionality after review" means someone has to understand what the code does before it ships. That's the part vibe coding skips.

Every App Store, every enterprise, every open source project is going to need this rule. Apple just wrote it down first.

[Ryan Loots @_rloots]

Apple isn't blocking AI. They're enforcing the governance layer that every platform will eventually need.

Their rule is simple: an app must not change its functionality after review. Vibe-coded apps break this because the developer doesn't know what functionality they shipped.

This is what happens when generation outpaces comprehension. Apple's just the first gatekeeper to say it out loud. Enterprise engineering orgs are next.

[Ryan Loots @_rloots]

The hardest new skill isn't choosing when to stop. It's knowing what you shipped.

When designers ship code to production, who reviews it? Not other designers, they don't read code. Not engineers, they're already drowning in AI-generated PRs.

You just added a new class of contributor with zero governance infrastructure around them. Same pattern everywhere: execution got cheap. Control didn't keep up.

[Ryan Loots @_rloots]

@wordgrammer The 30 minutes of real work becomes the most consequential 30 minutes in the company. Scoping what agents can touch, catching model drift before it hits production, containing blast radius when it doesn't. The job shrinks in hours and expands in stakes.

[Ryan Loots @_rloots]

@rohit4verse The 5% isn't systems thinking in the abstract. It's four specific skills: eval design, agent identity governance, behavioral versioning, blast-radius containment. That's the new senior engineer job description. It just doesn't have a title on any job board yet.

[Ryan Loots @_rloots]

@Pirat_Nation Linux just did what no enterprise has: made humans explicitly accountable for AI-generated code. The person who submits it owns it, including the governance, the tests, and the intent. That's the entire missing layer.

[Ryan Loots @_rloots]

https://t.co/EkgbFLL32I

[Ryan Loots @_rloots]

@hasufl Same pattern in engineering. AI agents now outnumber engineers 144:1 in some orgs and more than half run without security oversight or logging. The cost of producing governance artifacts went to zero, but so did the cost of ungoverned action at production scale.

[Ryan Loots @_rloots]

@theo The mid-tier role didn't disappear, it mutated. The new version builds eval suites, scopes agent identities, versions behavioral dependencies. Nobody's hiring "mid-tier developer" because the job title changed before the job boards caught up.

[Ryan Loots @_rloots]

@nayibbukele It's worse than a demand collapse. The people being cut are the ones who understand why the systems work. Companies are destroying the knowledge base that makes their own AI-built infrastructure maintainable. You can't govern what you can't explain.