Videos

Binny's take on Safe Generative AI on Avory & Co.

Name: Binny's take on Safe Generative AI on Avory & Co.
Uploaded: 2026-04-08
Duration: 29 min 37 s

Binny Gill on the Avory & Co. Inside Scoop podcast: English-as-code, the difference between creative AI (Concierge) and diligent AI (the Brain), and why humans must keep the steering wheel.

About this episode

Sean Emery from Avory & Co. hosts Kognitos founder and CEO Binny Gill on the Inside Scoop podcast for a 30-minute conversation on generative AI, the path from low-code to English-as-code, and what “safe” generative AI actually means in practice.

Five ideas Binny argues for

Only 25M people speak machine: ~25 million developers gate all of software innovation today. Generative AI breaks that bottleneck — when machines understand English faithfully, anyone with an idea can build.
English as code, not no-code: low-code and no-code platforms hide complexity behind menus, which limits power. English-as-code keeps every nuance the business cares about visible and editable.
Ambiguity is now a feature: the same ambiguity that disqualified English as a programming language in the 60s is now its strength. The runtime asks clarifying questions when intent is unclear, learns the answer, and applies it forward.
Creative AI vs Diligent AI: Kognitos pairs Concierge (the creative side that designs and proposes) with the Brain (the diligent side that runs an auditable English program). Creativity for the brainstorm; diligence for the execution.
Keep the steering wheel: humans must read and approve the program before it runs. Don't hand APIs directly to a model and let it hallucinate its way through enterprise data.

Quotes worth highlighting

“We're in the Dark Ages — 1 in 200 people knows how to teach a computer something new, yet all of humanity depends on computers.”
“The biggest innovation we've done is exception handling and ambiguity resolution. The system is smart enough to ask you a question instead of throwing a syntax error.”
“There is a line between creativity and guesswork, between intuition and logic — that's the line we help enterprises define.”
“Be a good GPT-4 jockey. Everyone in my company is expected to be using GPT-4 — otherwise you're on a horse and everyone else is in a race car.”

How Binny views the next decade

Binny predicts that roles will shift from “doing the work” to “reviewing the work” — students will think more like teachers, individual contributors will think more like managers. Interviews will test for the quality of questions, not just answers. The language-model layer itself becomes commoditised (like the electricity grid) — the winners are the businesses that use it to solve real problems with a human still in the loop.

Questions answered in this video

What's Binny's argument for “English as code” over low-code or no-code?

Low-code and no-code hide complexity behind menus, which limits the power of the platform. English-as-code keeps every nuance the business actually cares about visible and editable — the human and the machine are reading the same program at the same time.

What are the “two AIs” inside Kognitos?

Concierge — the creative AI — designs, proposes, and brainstorms automations conversationally. The Brain — the diligent AI — executes the resulting English program in a deterministic, auditable runtime. Creativity for design, diligence for execution.

Why did Binny say ambiguity in English is now an advantage?

In the 60s and 70s, ambiguity disqualified English as a programming language — too many possible interpretations. Today's runtimes can ask the user a clarifying question, learn the intended meaning, and apply it on every future run, turning the old weakness into a strength.

What's the “steering wheel” argument?

Binny argues that humans should always be able to read what a machine is about to do before it does it. Giving an AI direct API access and letting it act without a readable plan removes the steering wheel — and humans don't tolerate not being in control of consequential systems.

How does Binny see jobs changing in the next 10 years?

He frames it as an “up-levelling”: roles shift from doing work to reviewing work. Students will think more like teachers, ICs more like managers, and interviews will probe the quality of questions rather than rehearsed answers — which is more cognitively taxing, not less.

END