Development history
PublicVoiceShield evolved through product research, problem definition, rapid implementation and iterative refinement. The milestones below are grouped by month rather than presented as a commit-by-commit timeline.
Pre-development concept phase
- Identified the impact of hostility and trolling on public participation
- Defined the need to distinguish criticism, disagreement, abuse and genuine questions
- Established the principle: “AI advises. Humans decide.”
- Developed the idea of serving public figures, organisations, supporters and members of the public
- Outlined the responsibilities of handling public commentary safely and lawfully
- Set out the expectation that every stored record would be tenant-isolated from the outset
June 2026
- Early product research and decision-framework development
- Exploration of public-figure communication, audience risk and responsible reply support
- Definition of human oversight, escalation and audit requirements
- Data modelling work: deciding what needed to be stored, for how long, and who should be able to see it
- Design of the multi-tenant database structure around organisations and their members
- Planning of the append-only audit approach so decisions could be evidenced rather than overwritten
July 2026
- PublicVoiceShield MVP implementation
- Database installation and configuration on managed PostgreSQL
- Creation of the core schema: profiles, organisations, organisation members and invitations
- Creation of the analysis schema: social posts, thread analyses, analysed comments and comment decisions
- Creation of the research and evidence schema: research sessions, queries, sources, cache and evidence approvals
- Creation of the governance schema: privacy requests, policy versions, policy acceptances, cookie preferences and retention settings
- Row-level security policies, explicit table grants and security-definer membership checks
- Database indexing and performance review for higher comment and thread volumes
- Authentication and organisation accounts
- Analyse Thread workflow
- Secure server-side AI analysis
- Reply identity modes
- Human approval and audit trail
- Comment Review workspace
- Optional public-source research
- UK-law-assisted replies
- Privacy, legal and compliance pages
- Retention wording corrections
- Security and scalability review
- Tenant-isolation verification across organisations
- Tutorial video
- Private source-control integration
- Published controlled MVP
Created and developed by Mohammad Asif. Built from sustained product thinking, rapid implementation and iterative testing.