What should an engineer's core skill be once agents write most of the code?

Joelle Pineau · Nov 3, 2025

As code generation becomes abundant, the scarce and valuable function becomes selection and editorial judgment about which code is actually worth running

By analogy with image generation today, where volume is huge and what matters is picking quality out of the volume; someone still has to decide what the digital world should be running

Scope: extrapolated from the image-generation analogy

38:31 20VC: Cohere's Chief AI Officer on Why Scaling Laws Will Continue | Whether You Can Buy Success in AI with Talent Acquisitions | The Future of Synthetic Data & What It Means for Models | Why AI Coding is Akin to Image Generation in 2015 with Joelle Pineau

Joelle Pineau · Nov 3, 2025

Curation and verification are work that does not go away

39:14 20VC: Cohere's Chief AI Officer on Why Scaling Laws Will Continue | Whether You Can Buy Success in AI with Talent Acquisitions | The Future of Synthetic Data & What It Means for Models | Why AI Coding is Akin to Image Generation in 2015 with Joelle Pineau

Joelle Pineau · Nov 3, 2025

Humans remain necessary for intent and critique even in a world of mass-generated code, but team composition changes significantly

Someone must decide what to build and what purpose it serves, and designers who can go directly from idea to digital artifact change the team equation

39:45 20VC: Cohere's Chief AI Officer on Why Scaling Laws Will Continue | Whether You Can Buy Success in AI with Talent Acquisitions | The Future of Synthetic Data & What It Means for Models | Why AI Coding is Akin to Image Generation in 2015 with Joelle Pineau

Maor Shlomo · Nov 24, 2025

Within two years 95-100% of code at companies will be AI-written, with engineers mostly prompting agents

Base44 is already close to 90% because the repository was structured in the age of LLMs to be easy for models to navigate and read

58:17 20VC: Base44's Maor Shlomo on How Vibe Coding Will Kill SaaS and Salesforce | Why it is BS that Vibe Coding Platforms Do Not Have Defensibility and Bad Margins | Why He Worries About Google, Not Replit and Lovable | Why Long Anthropic, Not OpenAI?

Adam Foroughi · Apr 27, 2026

AI-native engineers don't need to write code but must be imaginative product thinkers who can audit code and know the business KPIs they're optimizing for.

You can't type out what you need in a complex system and get a finished deliverable, so engineers must review code for safety and quality while aligning agent output with measurable business value.

31:17 20VC: Applovin: $160BN Market Cap, $5.48BN Revenue, $10M EBITDA Per Head | Why the Best Do Not Need Mentorship | Why Founders Should Not Angel Invest | Why Kindness in Business Will Slow You Down with Adam Foroughi

David Luan · Jun 24, 2024

AI will make humans at work much more generalist, with individuals holding ever larger scope across functions while supervising a cohort of specialist AI copilots

Collapsing multiple skill sets into one person makes work move faster, and AI copilots can supply the specialist depth

39:14 20VC: Why Foundation Model Performance is Not Diminishing But Models Are Commoditising, Why Nvidia Will Enter the Model Space and Models Will Enter the Chip Space & The Right Business Model for AI Software with David Luan, Co-Founder @ Adept

Anton Osika · Aug 18, 2025

The skills of a good engineer shift toward generalism, with AI supplying the deep expertise

Being a generalist lets you understand how things come together as a larger whole, which matters more when AI can cover depth

36:30 20VC: Lovable CEO Anton Osika on $120M in ARR in 7 Months | The Honest Truth About Defensibility and Unit Economics for AI Startups | The State of Foundation Models: Long Grok, Short OpenAI, Why | Replit vs Lovable vs Bolt: What Happens

Alexandr Wang · Jun 12, 2024

Software engineering will change dramatically, with the durable, valuable skill becoming translation of customer problems into scoped engineering tickets that an AI engineer can solve

Models will get better and better at coding, so developers won't need to spend time on much of what they do today, but parts of the job remain irreplaceable

29:10 20VC: Scale's Alex Wang on Why Data Not Compute is the Bottleneck to Foundation Model Performance, Why AI is the Greatest Military Asset Ever, Is China Really Two Years Behind the US in AI and Why the CCPs Industrial Approach is Better than Anyone Else's

Alexander Embiricos · Feb 21, 2026

The spec or plan for what you want built is more important than ever, and reviewing the plan is becoming more important than reviewing the code

We are entering a delegation phase of working with agents, so agreeing architecturally up front — like a new hire presenting an RFC before starting work — is where the leverage is

Scope: framed as an underrated thing, alongside, not instead of, code review

21:42 20VC: Codex vs Claude Code vs Cursor: Who Wins, Who Loses | Will All Coding Be Automated - Do We Need PMs | The Real Bottleneck to AGI | The Three Phases of Agents and What You Need to Know with Alex Embiricos, Head of Codex at OpenAI

Maor Shlomo · Nov 24, 2025

With AI builders you should throw away your first version rather than write a PRD and plan ahead — fast iteration and hitting revert beats planning

Planning existed to save developer time, which no longer applies; once you've figured out the product through iteration you can reach the same place in a few prompts, and the model will suggest features you hadn't considered

59:04 Throwaway iteration replaces upfront specs

Martin Casado · Jul 28, 2025

AI is eliminating the middle of the software value chain — the boilerplate — while the frontier of new computer science and the go-to-market long tail remain human

The average production pull request is ~2 lines; the hard part is the field learning that produces those lines, not writing them

40:38 Ai eliminates the middle boilerplate leaving frontier cs and gtm to humans

Raaz Herzberg · Dec 12, 2025

Core engineering fundamentals like algorithmic thinking will remain valuable no matter how coding is abstracted away by AI

Python was already a big abstraction over low-level languages; what enables you to build complex systems and debug is the underlying concepts, which she says even help her in marketing

63:47 Algorithmic fundamentals survive every abstraction layer

Matan Grinberg · Jun 13, 2026

What makes a great engineer changes materially in the agent era: the best engineers own full end-to-end business outcomes spanning product, sales enablement and marketing rather than just shipping features

Those are the parts of engineering that have historically made engineers good founders, and they're what actually matters when the coding itself is delegated

28:33 End to end business outcome ownership across product sales and marketing

Matan Grinberg · Jun 13, 2026

The future of software organizations is engineers who build the factories that build the software rather than engineers who write the code

Like Tesla's factory, humans design the process to optimize throughput even if few humans are on the assembly line; humans ensure output isn't bloat or debt-accruing test-passing code

38:28 Engineers design and operate the agent factory