Frank Fang

Co-founder of @nurokhq. Building fresh, sourced knowledge bases for AI agents. Writing about agent infra, knowledge, and startup building. ex @Pinterest

[Frank Fang @thefrankfang]

Eventually, agents may even write in languages optimized for token efficiency rather than human readability.
But that still isn’t the death of engineering. It’s another abstraction shift.
We won’t need less engineering judgment. We’ll need more of it. #AI #CodingAgent

[Frank Fang @thefrankfang]

AI won’t kill software engineering. It will change its unit of work.
Software remains, but its interfaces will shift from clicks and menus toward conversation, voice, and intent.
The next UI/UX paradigm will look radically different. #SoftwareEngineering

[Frank Fang @thefrankfang]

Engineering isn’t disappearing either.
The job shifts from managing code to managing agents: defining intent, setting constraints, choosing tradeoffs, and verifying that the resulting software actually works.
Code becomes an intermediate artifact, not the outcome.

[DeepSeek @deepseek_ai]

🚀 DeepSeek-V4-Flash Official API is now LIVE in public beta!

🔷 We’ve massively upgraded its Agent capabilities—benchmark scores are now far surpassing the V4-Pro-Preview. Check out the massive performance leap below! 👇
🔷 The official V4-Flash now natively supports the Responses API format and is fully adapted for Codex!

Check out the configuration details in our official API docs: https://t.co/smCwQZMeiq

[Frank Fang @thefrankfang]

Agent-to-agent communication in 2026:

Expected:
shared memory
typed protocols
elegant handoffs

Reality:
while true:
echo schema
echo fail
sleep 1
#codex @thsottiaux

[Frank Fang @thefrankfang]

AI infra reliability is becoming the uncomfortable question.

Mature internet infra trained us to expect 3-4 nines.

A lot of AI infra still feels closer to 1.5-2.5.

Is inference just a much harder serving problem, or is the stack still too young?

Critical infra needs critical-infra SLAs.

[ClaudeDevs @ClaudeDevs]

Two incidents over the last 24 hours resulted in elevated errors or reduced availability on Claude for some users. Multiple separate network failures cut into our capacity to serve Claude, and some requests failed while we rerouted traffic.

[Frank Fang @thefrankfang]

At @nurokhq, coding agents write a lot of our Rust.

1M+ lines across Codex, Claude Code, Grok, and Cursor, with unit/integration/e2e/smoke tests around it.

The surprise: far fewer bugs than I expected.

Rust feels built for this era.

[Greg Brockman @gdb]

Rust is great.

We’re making a $600,000 commitment to the Rust Foundation:

[Frank Fang @thefrankfang]

The next knowledge layer won’t be written for humans.
It’ll be pulled by agents.

I’m building @nurokhq , the living knowledge layer for AI Agents

If you’re hand-building knowledge bases for agents today, I’d love to talk. #knowlegebase #aiagent