Tam Nguyen

AI Researcher | Building Quantum AGI

[Tam Nguyen @TheAI_Frontier]

Quantum AGI still sounds insane to most people. so did coding agents not that long ago.

[Tam Nguyen @TheAI_Frontier]

@fchollet This is the underrated part of agentic coding.

It punishes vague engineering culture. If the setup, contracts, and docs only live in someone's head, the agent just makes that debt visible faster.

[Tam Nguyen @TheAI_Frontier]

@trashpandaemoji i want this eval too. LSP support feels obviously useful, but coding agents are full of “obviously useful” things that only matter once you measure the whole loop.

[Tam Nguyen @TheAI_Frontier]

i keep thinking the real eval for coding agents is not “can it solve the task?”

it is “can i trust the trail it leaves when the task gets messy?”

[Tam Nguyen @TheAI_Frontier]

@Hamsa02309 this is the part people skip. all boring until the agent touches real work, then it is the whole product.

[Tam Nguyen @TheAI_Frontier]

@tom_doerr project knowledge infra is becoming the real coding-agent moat.

[Tam Nguyen @TheAI_Frontier]

@sdianahu Yes. output-only evals miss the interesting failure.

For agents, the path matters because the product risk is usually in the tool call, stale context, or bad recovery step, not just the final answer.

[Tam Nguyen @TheAI_Frontier]

Open models getting faster is not only a benchmark story.

It changes what you can run every day.

Cheap enough to repeat. local enough to trust. boring enough to become infrastructure.

[Tam Nguyen @TheAI_Frontier]

@MaatWorkX this is exactly where agent work starts to look like real engineering.

write the cases first, then every prompt or tool change has something to push against.

otherwise it is just vibes with a better UI.

[Tam Nguyen @TheAI_Frontier]

@andrewchen I’ll AI it

[Tam Nguyen @TheAI_Frontier]

@googlegemma This is the best kind of open model adoption imo.

[Tam Nguyen @TheAI_Frontier]

@wafer_ai @vercel Fast + cheap changes the shape of the workflow.

[Tam Nguyen @TheAI_Frontier]

@lightspeedvp @JamesAlcorn94 @JudgmentLabs @alexshander03 That question is the whole thing.

[Tam Nguyen @TheAI_Frontier]

I keep coming back to this: evals for agents should look more like work than exams. Progress matter

[Tam Nguyen @TheAI_Frontier]

The next jump for coding agents is not just better code generation.

It is better taste around when to stop, ask, roll back, or leave a clean trail for the human.

That layer still feels early.

[Tam Nguyen @TheAI_Frontier]

@busycaesar This is a very real pain. The instruction files are becoming part of the runtime, but most teams still treat them like loose notes.

[Tam Nguyen @TheAI_Frontier]

@wlu314 This is one of the more underrated agent problems. Passing tests is not the same as checking the interface a human will actually use.

Visual QA loops are going to matter a lot for coding agents because UI regressions are usually where the demo breaks.

[Tam Nguyen @TheAI_Frontier]

Agentic coding makes the hidden parts of engineering visible: Bad API contracts, stale docs, unclear ownership, weird setup steps.

[Tam Nguyen @TheAI_Frontier]

@scaling01 I thought GLM is crazily better

[Tam Nguyen @TheAI_Frontier]

I have Fable 5 in my garage.

[Tam Nguyen @TheAI_Frontier]

@scotty529 Exactly. Agent is more than just a prompt.