Hoang Le

Engineering leader & founder. Helping founders go from zero to live users. Writing about engineering leadership, product development, and building with clarity.

[Suryansh Tiwari @Suryanshti777]

Andrej Karpathy just explained the future of software engineering without directly saying it.

The best AI engineers are no longer “prompting.”

They’re building systems around the agents.

Karpathy’s biggest insight wasn’t:
“Claude can code.”

It was:
LLMs become dramatically better when you force them into disciplined workflows.

That’s why "CLAUDE.md" files are suddenly everywhere.

Not because they’re prompts.

Because they behave like an operating system for the agent.

Karpathy called out the exact problems with AI coding:

- models assume instead of asking
- they overengineer simple tasks
- they hide confusion
- they rewrite unrelated code
- they optimize for completion, not correctness

So developers started encoding rules directly into the workflow:

→ Think before coding
→ Simplicity first
→ Surgical edits only
→ Goal-driven execution

And the results are wild.

People are now running multiple Claude Code agents in parallel like engineering teams:

• one agent researching
• one debugging
• one writing tests
• one optimizing code
• one validating outputs

Not “AI assistance.”

Actual orchestration.

And this part from Karpathy changes everything:

“Don’t tell the model what to do. Give it success criteria and let it loop.”

That is the shift.

From:
“write this function”

To:
“here’s the goal, constraints, tests, and verification system — now iterate until correct.”

The craziest part?

This already feels like a phase shift in engineering.

A lot of developers quietly went from:
80% manual coding → to 80% agent-driven coding in just months.

Not because AI became perfect.

Because the leverage became impossible to ignore.

We’re entering an era where the highest leverage engineers won’t necessarily be the best coders.

They’ll be the people who build the best systems around AI agents.

[Jaana Dogan ヤナ ドガン @rakyll]

I'm not joking and this isn't funny. We have been trying to build distributed agent orchestrators at Google since last year. There are various options, not everyone is aligned... I gave Claude Code a description of the problem, it generated what we built last year in an hour.

[Hoang Le @hoangleitvn]

15 years building software. 50+ projects. Same pattern:

Teams ship fast. Then slow down.

Not because they got lazy. Because the foundation was never there.

I started writing about the 80% nobody talks about.

https://t.co/WpZ5M6lsqU

[Hoang Le @hoangleitvn]

@karpathy The skill gap isn't about mastering agents, MCP, or the tool alphabet soup. It's a mental shift. I think AI adoption is a workflow decision, not a tool decision. You need to change how you work with AI.

[Liad Yosef @liadyosef]

Huge thanks to @github for the amazing shout-out on MCP-UI and the new MCP Apps spec!

We're proud to join forces with @OpenAI and @AnthropicAI to create a unified spec for apps that run across chat platforms.

Build once, run everywhere. 🚀

(cc @idosal1 )

[Kent C. Dodds 🏹 @kentcdodds]

I've got over 200 blog posts and pages powered by MDX hosted on GitHub for my personal blog and it's been great!

It's compiled on demand and cached in SQLite and globally distributed (not at the document level because every page is server rendered with dynamic content).

[Hoang Le @hoangleitvn]

Full breakdown with pricing tables and decision trees:

https://t.co/yi9Z6AMje0

Running a t3 fleet? What's your migration plan?

[Hoang Le @hoangleitvn]

Closing my AWS re:Invent 2025 deep dive.

Here are the database cost updates that actually matter for 2026:

[Hoang Le @hoangleitvn]

Counterintuitive: for always-on dev environments, staying on t3 with Reserved Instance is cheaper than upgrading to t4g.

Newer isn't always cheaper when discounts don't exist.

[Hoang Le @hoangleitvn]

8/ Pricing reality. EKS control plane is $0.10 per cluster-hour in standard support, $0.60 in extended support. Capabilities add pay-per-managed objects, so you need a quick cost model vs current compute and ops time.

[Hoang Le @hoangleitvn]

EKS updates worth watching.

EKS Capabilities now manages core building blocks (Argo CD, ACK, KRO). Auto Mode keeps getting smarter. Kubernetes lifecycle rules are clearer, so upgrades are easier to plan.

https://t.co/3Fi7CkKUFd

[Hoang Le @hoangleitvn]

7/ AWS guidance in practice. Use Capabilities when overhead reduction and managed lifecycle matter. Self-manage when you need deep customization or specific controller behavior.