Manifesto for AI‑Driven & AI‑Assisted Development
A call to build tools that lift the whole economy—not just the lab bench
Preamble
Artificial Intelligence is already shifting from a curiosity of research labs to a universal force for broad‑spectrum automation. The greatest value will not come from a handful of virtuoso models chasing Nobel‑level breakthroughs, but from systems that quietly—but relentlessly—take drudgery out of everyday work, amplify human creativity, and free minds for higher‑order problems.
This manifesto sets out the principles that should guide teams, companies, and communities as we design, deploy, and govern AI‑driven and AI‑assisted solutions.
We have come to value …
More valuable
- Broad automation of routine work
- Human‑AI symbiosis
- Iterative, real‑world deployment
- Outcome‑based metrics (productivity, well‑being, equity)
- Transparent, auditable systems
- Continuous up‑skilling & job redesign
Less valuable
- Niche optimization of boutique research tasks
- Pure substitution of people with models
- "Big‑bang" unveilings in isolated sandboxes
- Model‑centric benchmarks divorced from use‑cases
- Proprietary black boxes
- One‑off retraining after layoffs
(While there is value on both sides, items on the left should guide our priorities.)
Ten Principles
1. Automate Where the Money—and the Drudgery—Is
Start with the high‑volume, low‑joy tasks that eat real budgets and worker hours. Every spreadsheet reconciled or help‑desk ticket triaged is a step toward economy‑wide uplift.
2. Design for Symbiosis, Not Supremacy
Tools should extend human agency—drafting, checking, summarizing, and coordinating—while leaving judgment, empathy, and vision to people.
3. Ship Fast, Learn Faster
Incremental releases in production settings trump perfect pilots. Measure impact on throughput, error rates, and user satisfaction, then refine.
4. Treat Data and Compute as Public Infrastructure
Open standards, shared datasets, and federated compute keep progress from bottlenecking on proprietary silos.
5. Build for Multimodal Reality
R&D isn't only code and papers; neither is the rest of work. AI must see, hear, manipulate, and coordinate across physical as well as digital domains.
6. Embed Safety & Ethics at the Architecture Level
Guardrails, provenance tracking, and red‑team stress‑tests belong in CI/CD pipelines, not tacked on during compliance reviews.
7. Prioritize Economic Inclusion
Redeploy productivity gains to reskill and redeploy workers long before displacement becomes unemployment.
8. Measure Real‑World Externalities
Carbon, supply‑chain strain, and social trust are first‑class KPIs. If a model's energy bill outweighs its economic value, rethink the approach.
9. Govern Through Radical Transparency
Publish model cards, decision logs, and usage dashboards. Scrutiny—by regulators, users, and affected communities—is a feature, not a threat.
10. Think in Decades, Act in Sprints
Broad automation is a marathon of compounding gains, punctuated by short cycles of invention. Keep horizon and cadence in healthy tension.
Call to Action
Whether you are a lone developer scripting workflows, a product lead integrating LLMs, or a policymaker drafting guardrails, commit to these principles. Aim for the wide lens, where automating 1,000 mundane tasks creates more value—and more human opportunity—than automating one moon‑shot experiment.
The future we build with AI will be measured not by how many papers it publishes, but by how many lives it quietly makes better.