Funniest thing from the game. Arsenal got a corner kick. They were taking their usual 15 minutes to crowd the goalkeeper. The referee said that’s enough and blew the half time whistle.
😂😂😂😂😂😂
We’re moving beyond “AI chatbots” into a new stack: Agentic UI.
Architecture combines:
• Generative UI → agents dynamically create interfaces
• State synchronization → frontend + agent stay consistent in real time
• Interrupt-driven approvals → humans step in for actions
• Event streams → agents continuously update the UI as they reason
The important shift:
👉 agents are no longer just generating text
👉 they’re becoming active participants in application state and workflows
Interesting pattern across AI engineering:
The teams getting the best results from coding agents are not the teams with the best prompts.
They’re the teams with:
• strong architecture
• clean repositories
• clear task boundaries
• good verification loops
• context management
The AI industry may be hitting its first real economic constraint.
When even Microsoft reportedly cancels internal Claude Code licenses over token costs, it signals something important:
👉 the “AI is getting infinitely cheaper” narrative is colliding with production reality.
Agentic workflows consume enormous compute because they:
• maintain long contexts
• call tools repeatedly
• run verification loops
• generate/refactor large codebases
• operate continuously, not interactively
This breaks the economics of traditional SaaS pricing.
Flat-rate subscriptions worked when humans typed prompts occasionally.
But autonomous agents create machine-scale demand against GPU-bound infrastructure.
The industry is now confronting a hard question:
Can frontier AI sustain:
• high usage
• low pricing
• and healthy margins
…all at the same time?
This is why the next wave of innovation may focus less on “smarter models” and more on:
• context compression
• caching
• smaller specialist models
• local inference
• efficient orchestration
• token-aware architectures
The future winners may not be the companies with the biggest models.
They may be the ones with the best economics per intelligent action. ⚙️🤖
A strong AI engineer doesn’t just learn frameworks — they build mental models.
Some standout books from Javarevisited’s AI Engineering reading list:
📘 AI Engineering — Chip Huyen
📘 Designing ML Systems — Chip Huyen
📘 The LLM Engineering Handbook — Paul Iusztin & Maxime Labonne
📘 Build a Large Language Model (From Scratch) — Sebastian Raschka
📘 Hands-On Large Language Models — Jay Alammar & Maarten Grootendorst
📘 Building LLMs for Production — Louis-François Bouchard & Louie Peters
The progression is clear:
LLM fundamentals → production systems → agentic workflows → evaluation & reliability.
The future AI engineer won’t just prompt models.
They’ll understand the full stack:
data, retrieval, orchestration, infra, memory, and deployment. 🧠⚙️ #AIEngineering
“Agentic programming” is emerging as a new engineering discipline.
Not just prompting models — but designing systems where AI can:
• plan tasks
• use tools
• manage memory
• recover from failures
• operate in long-running loops
The roadmap is becoming clearer:
prompt engineering → context engineering → workflow orchestration → autonomous systems.
The key shift:
👉 building AI apps is starting to look less like frontend development
👉 and more like distributed systems engineering.
Future AI engineers won’t just write code.
They’ll design environments where agents can reliably reason and act. ⚙️🤖
We are investigating unauthorized access to GitHub’s internal repositories. While we currently have no evidence of impact to customer information stored outside of GitHub’s internal repositories (such as our customers’ enterprises, organizations, and repositories), we are closely monitoring our infrastructure for follow-on activity.