TwoSetAI

[TwoSetAI @TwoSetAI]

"Hire better people, pay them more money, and have fewer of them." When code turns into plug and chug, headcount stops being the lever it used to be.

Full video on TwoSetAI@Youtube. @hi_angelinayang

[TwoSetAI @TwoSetAI]

AI didn't shorten the workday — it raised the bar for what "shipped" means. Tabs CTO Deepak Bapat on why AI lets teams do more with more, not just more with less.

Full video on TwoSetAI@Youtube. @hi_angelinayang

[TwoSetAI @TwoSetAI]

"Great products, very very thin technology." SaaS made underengineering survivable. That window closed — and product companies with no real tech are the ones exposed.

Full episode on the TwoSetAI channel. @hi_angelinayang

[TwoSetAI @TwoSetAI]

Junior devs are "wizards casting spells" they don't understand, says John Berryman. AI agents never stump their toes learning syntax — so when things break, no one knows how to debug. "I think we're all doomed."

Full video on Youtube: TwoSetAI. @hi_angelinayang

[TwoSetAI @TwoSetAI]

Hallucination isn't the model making things up. It's picking wrong from a set that's too big. Shrink the output space and it becomes a selection error.

Full video on Youtube: TwoSetAI. @hi_angelinayang

[TwoSetAI @TwoSetAI]

John Berryman set up Cursor to know it's working with his 8-year-old. Every game the kid finishes auto-publishes to his own arcade site. He's made about 100 so far.

Full video on Youtube: TwoSetAI. @hi_angelinayang

[TwoSetAI @TwoSetAI]

"I saw it on TV. It's real." Daniel Davis on why this 90s Wag the Dog line hits harder than ever in the AI era.

Now on YouTube: TwoSetAI @hi_angelinayang

[TwoSetAI @TwoSetAI]

@clickup @Alexclickup @mathemagic1an Thanks @hi_angelinayang

[TwoSetAI @TwoSetAI]

Why assign the same task to multiple AI agents? Because LLM randomness = diversity, not noise. Simon Du compares it to bootstrap sampling — more coverage through cooperation, not competition.

Full video in comment. @hi_angelinayang

[TwoSetAI @TwoSetAI]

Prompting your AI more ≠ more control.

Tabs CTO Deepak Bapat on why letting users prompt on top of a trained context object actually causes MORE hallucinations, not fewer.

Full video link in comment. @hi_angelinayang

[TwoSetAI @TwoSetAI]

Why pay $100/mo for a product that's 80% right?

@JnBrymn John says soon you'll just tell an AI "take this, change that" — exactly-what-you-want software.

That's hyper-personalization, and it's a whole new abstraction level.

Full video link in comment. @hi_angelinayang

[TwoSetAI @TwoSetAI]

"The whole reason they hired me is to blame someone when things go wrong."

Tabs CTO Deepak Bapat on why AI won't fully replace finance teams — liability doesn't scale like software.

Full video link in the comment. @hi_angelinayang

[TwoSetAI @TwoSetAI]

"Fred has four legs" isn't just a fact — it's a statement. "Mark told me on Friday" is a statement about that statement. That's reification.

Daniel Davis breaks it down. Full video coming soon. Stay tuned. @hi_angelinayang

[TwoSetAI @TwoSetAI]

Search "gorilla suit," find "monkey costume" — that's semantic search understanding ideas, not just words.

But it can't filter by price or color like old exical search can. John Berryman (ex-GitHub Copilot) on when to use which.

Full video coming soon. @hi_angelinayang

[TwoSetAI @TwoSetAI]

"If we let people prompt on top of it, you're grading real trust." Tabs CTO Deepak Bapat on why his AI finance platform has NO chat interface — on purpose.

Full video coming soon, stay tuned 👀 @hi_angelinayang

[TwoSetAI @TwoSetAI]

No AI model holds 10M+ tokens of context. ApodX's fix: divide and conquer. Break the problem into simple sub-tasks — each agent only needs a sliver of context. Simon Du explains how 150 agents pull it off.

Full video coming soon. Stay tuned. @hi_angelinayang

[TwoSetAI @TwoSetAI]

LiveKit's Tom Shapland: AI voice agents can start "thinking" of a reply before you even finish talking — just like human brains predict what someone will say mid-sentence. 🎙️

Full video coming soon. Stay tuned. @hi_angelinayang

[TwoSetAI @TwoSetAI]

@livekit 's Tom Shapland: "The toothpaste isn't going back in the tube." AI coding agents are here to stay — but engineering judgment isn't going anywhere either. 🎙️

Full video coming soon. Stay tuned @hi_angelinayang

[TwoSetAI @TwoSetAI]

"It's easier to organize 100 agents than 100 humans." Simon Du (ApodX) on why LLMs are actually more predictable than people — and how verification + redundancy stop errors from compounding across agent teams.

Full video coming soon. Stay tuned. @hi_angelinayang

[TwoSetAI @TwoSetAI]

@livekit 's Tom Shapland: stop letting AI coding agents be "the blind leading the blind." Clone the open source repo first — let the AI reference real code, not stale training data. 🎙️

Full video coming soon. Stay tuned. @hi_angelinayang