Why I’m writing about this
I’ve been watching the claim rippling through engineering circles: that the act of writing code by hand is becoming a solved engineering task. The person who made the boldest, most concrete version of that claim is Boris Cherny boris@anthropic.com, the creator and head of Claude Code at Anthropic. He frames the moment as software’s "printing‑press" event: as tooling turns code into a commodity, the scarce resource shifts toward domain knowledge and product judgment. I want to unpack what he meant, the evidence he and reporters cite, and what every builder should be preparing for.
Summary of the thesis
Boris Cherny boris@anthropic.com argues that for a very large slice of modern software (web stacks, standard APIs, common data transformations), models like Claude can already generate, test, and refactor code reliably enough that human typing is optional. That shifts the bottleneck: instead of needing people who can type code, you need people who know what to build, how users think, and how to orchestrate AI agents that implement the idea.
Evidence and examples he uses
- Inside Anthropic, the Claude Code team reports dramatically higher per‑engineer productivity; anecdotal examples include dozens to hundreds of model‑authored pull requests per day and teams where non‑engineers ship code because agents handle the mechanical work.WorkOS and other writeups document these internal metrics.
- Public indicators: analyses showing accelerating shares of AI‑authored commits on GitHub and company memos setting targets for AI‑generated code usage (multiple firms reported high AI‑generation percentages in recent coverage).Times of India
- Demonstrations of agent loops and orchestration: interviews and talks where Boris Cherny boris@anthropic.com describes running many autonomous agents, scheduled loops, and Cli‑first workflows that sidestep traditional IDEs.TeamDay
Taken together, the logic is straightforward: if models reliably produce correct code for common tasks, then the market can support many more people who direct those models — and those people don’t all need formal CS training.
What this implies for education and the labor market
- Curriculum pivot: a premium on domain expertise (accounting, logistics, health), systems thinking, data literacy, product design, and the ability to evaluate and orchestrate model outputs. Traditional CS fundamentals still matter for edge cases and system reliability, but teaching should move toward "how to scope and verify agentic solutions."
- Lower barrier to building: lower startup costs and faster iteration mean more niche products and more founders from non‑technical backgrounds. Cherny predicts a multiplication of builders; I expect many will be domain experts who learn to be product‑focused orchestrators.
- Labor churn and retraining: entry‑level developer roles may shrink or shift; mid‑career engineers who double down on systems, safety, platform engineering, and architecture will be valuable. Policies and companies must invest in retraining or risk large displacement.
Impact on engineering practices and tools
- Tooling evolves: IDEs may be reshaped into agent orchestration consoles, and CLI/loop paradigms may replace line‑by‑line editing for many flows. The emphasis moves from typing code to managing workflows, tests, and deployment logic.
- Review and QA change: first‑pass reviews may be automated, but verification layers and stronger testing/safety pipelines become the real differentiator. Engineers will spend more time defining correctness, invariants, and monitoring.
- Architecture and observability: systems must be built with agent interactions, provenance, and explainability in mind. Debugging becomes more about tracing agent intents and prompts than step‑through code execution.
What managers and engineers should do now
- Learn to orchestrate: invest in skills for prompt design, agent composition, and monitoring. If you’re a manager, create small projects that force teams to build with limited headcount and abundant model access — scarcity breeds better delegation patterns.
- Recenter hiring: hire domain experts and cross‑disciplinary generalists who can scope problems and own outcomes. Train existing staff on verification, safety, and systems thinking.
- Build verification culture: require rigorous tests, canarying, and observability for model‑generated outputs. Treat model outputs as first drafts that require domain validation, not black‑box production pushes.
- Invest in internal tooling and data hygiene: better internal APIs, datasets, and context make models more reliable. The organization that curates the best context and feedback loop will get disproportionate leverage.
Potential risks and downsides
- Overreliance and complacency: if humans stop understanding fundamentals, incidents in atypical edge cases will be harder to diagnose. Tacit knowledge can atrophy if we delegate everything too quickly.
- Concentration and monoculture: if a few models dominate, subtle biases or supply‑chain failures could cascade across many products. Diversity of models and open‑source alternatives matter.
- Job displacement and inequality: rapid automation risks concentrated harm for entry and mid‑level roles unless companies and governments plan for transition support and reskilling.
- Security and IP: agentic coding raises new attack surfaces — prompt injection, poisoned context, and IP provenance disputes will be common.
- Economic churn: faster startup formation is good for innovation but can destabilize incumbents and communities reliant on legacy roles.
Final thought — what I would do if I were building today
I’d start by teaching domain experts to become builders: give them model access, a small budget, and a sprint‑style mandate to ship. For managers, prioritize verification engineers and platform teams that capture context and iterate on model‑human loops. For individual engineers, invest in systems design, safety, and the ability to translate vague product needs into verifiable acceptance criteria — that skill will outlast the next architecture.
Boris Cherny boris@anthropic.com has given us a provocation and a roadmap. Whether you call the future "builders," "creators," or simply "people who direct agents," the practical question is the same: will we adapt our institutions — schools, hiring, and product processes — fast enough to make this a broad win instead of a concentrated disruption?
Regards,
Hemen Parekh
Any questions / doubts / clarifications regarding this blog? Just ask (by typing or talking) my Virtual Avatar on the website embedded below. Then "Share" that to your friend on WhatsApp.
Get correct answer to any question asked by Shri Amitabh Bachchan on Kaun Banega Crorepati, faster than any contestant
Hello Candidates :
- For UPSC – IAS – IPS – IFS etc., exams, you must prepare to answer, essay type questions which test your General Knowledge / Sensitivity of current events
- If you have read this blog carefully , you should be able to answer the following question:
- Need help ? No problem . Following are two AI AGENTS where we have PRE-LOADED this question in their respective Question Boxes . All that you have to do is just click SUBMIT
- www.HemenParekh.ai { a SLM , powered by my own Digital Content of more than 50,000 + documents, written by me over past 60 years of my professional career }
- www.IndiaAGI.ai { a consortium of 3 LLMs which debate and deliver a CONSENSUS answer – and each gives its own answer as well ! }
- It is up to you to decide which answer is more comprehensive / nuanced ( For sheer amazement, click both SUBMIT buttons quickly, one after another ) Then share any answer with yourself / your friends ( using WhatsApp / Email ). Nothing stops you from submitting ( just copy / paste from your resource ), all those questions from last year’s UPSC exam paper as well !
- May be there are other online resources which too provide you answers to UPSC “ General Knowledge “ questions but only I provide you in 26 languages !
No comments:
Post a Comment