Parliamentary Intelligence Layer
DailDex
Investor deck · 2026
DáilDex runs a custom fine-tuned AI agent harness — Hermes — that monitors all 160 Irish TDs continuously. A journalist opens the app and sees every vote, PQ, and speech from their TD this week, sourced and summarised, without doing a minute of research.
This used to require a newsroom. It now costs roughly €35 a week.
DáilDex — four core screens: Today's Dáil briefing, TD activity profile, party intelligence dashboard, AI political assistant.
160 AI agents. 160 TDs. Running continuously. No Irish product has ever watched the entire Dáil at once — until now.
The core technology
This is not a search engine over government PDFs. DáilDex runs a persistent agent layer that watches the Dáil in real time — processing every vote, every Parliamentary Question, every speech, and every party statement the moment it enters the public record. The agents never sleep, never miss a division, and never forget what a TD said six months ago.
TDs monitored simultaneously
Continuous surveillance — no breaks, no holidays
Lag from Official Record to structured data
Cross-TD memory — agents remember everything
The moment a division is called in Dáil Éireann, an agent processes the result, maps it to every TD's record, detects party-line breaks, and flags anything anomalous. No journalist could monitor 160 people at once. DáilDex does it continuously.
Parliamentary Questions are ingested, attributed, topic-classified, and linked to the submitting TD's issue history the moment they are published. Over time the agent builds a map of what each TD actually cares about — not what they say they care about.
Dáil speeches are long, full of jargon, and designed for the record rather than the reader. Hermes reads them, writes a plain-English summary, links it back to the transcript, and surfaces any position change relative to prior votes or statements.
Each party runs a dedicated AI room that remembers everything — positions, internal tensions, recurring topics, discipline breaches, spokesperson patterns. It is not a chatbot. It is a persistent analyst that never forgets and never sleeps.
The agent harness
DáilDex does not call a generic LLM API and hope for good summaries. We built a custom agent harness on top of the Hermes framework — fine-tuned on Oireachtas language, Dáil procedure, Irish political terminology, and parliamentary source citation patterns.
The result: 160 agents running in a persistent virtual world with their own browsers, their own memory, and the ability to contact TD offices directly. They don’t stop when you close the app. They don’t forget what they learned last month. They watch, they record, and they alert — autonomously.
Generic models hallucinate political context. Ours doesn’t — and ours can also send a message to the TD’s press office when the story breaks.
Hermes cron jobs run on natural-language schedules with no human oversight. The moment the Oireachtas publishes a vote, an agent wakes, processes it, and updates every TD profile — including at 3am on a Saturday.
Hermes ships full browser automation via CDP. Our agents browse oireachtas.ie, extract live transcripts, and pull Bills Office records directly — no API needed, no gate to wait behind.
Hermes has native outbound messaging across Telegram, Signal, WhatsApp, Slack, and Discord. DáilDex can send sourced accountability notifications directly to TD offices, press teams, and journalists.
Agents run inside HermesWorld — a persistent virtual environment with isolated operating zones. They continue working offline and surface structured updates autonomously, with no human trigger required.
Each TD gets a dedicated subagent instance. Hermes delegates concurrently — one per TD, one per topic, one per bill — with isolated context and restricted toolsets per agent. This is the architecture, not the limit.
A party agent that tracked a TD's housing position in January still knows it in October — no re-ingestion, no context loss. Every stance, vote, and speech is remembered and compared across the full political lifetime.
Market size
Democratic monitoring infrastructure is not a niche. Every parliament, every local council, every regulator in every English-speaking country has the same problem: public records, no intelligence layer. Ireland is the proof of concept.
Global political intelligence, civic data, and government monitoring software
English-speaking parliamentary democracies — UK, Ireland, Canada, Australia, NZ
Ireland B2B licensing, Pro subscriptions, and API deals — achievable in 3 years
Bottom-up: 160 TDs × institutional + pro + API revenue. Media API alone: 10 outlets × €1K/mo = €1.2M ARR, zero consumer dependency.
Competition
There are civic archives, a commercial monitoring tool, and official portals. Civic tools proved that demand for accessible parliamentary records exists — KildareStreet has been running since 2004. PolicyDesk proves institutional buyers pay for Irish political intelligence. Official portals are the raw data supply. None of them combines continuous agent monitoring, source-backed summaries, party memory, citizen UX, and institutional API products.
Archive search, alerts, and API over Dáil debates from 2004. The obvious civic predecessor.
Archive and search only. No AI summaries, no stance detection, no party memory, no institutional data products.
Unofficial explorer for Dáil/Seanad members, votes, speeches, and legislation using Oireachtas open data.
Clean record explorer. No agents, no alerts, no summarisation layer, no API monetisation.
Citizen voting alignment tool — compare your views with TD voting records. Plans bill summaries and gap scores.
Participatory/poll oriented. Narrower than DáilDex on coverage and not built around continuous monitoring.
The closest commercial competitor. Official-source political monitoring for Ireland — Oireachtas, government publications, and local authorities. B2B and public-affairs oriented.
No citizen accountability UX, no TD-level agent architecture, no party rooms or persistent memory. Monitoring reports, not explained intelligence.
Pulls from the Register of Lobbying. Links lobbying records to elected and public officials. 108,985 returns, 4,047 officials, 2,873 lobbyists.
Lobbying-specific only. DáilDex integrates lobbying as one layer inside a wider political intelligence graph.
The hard source of truth. Official debates, votes, PQs, bills, and members. Open APIs expose machine-readable records.
Not an intelligence product. No summaries, no agents, no monitoring, no alerts. DáilDex is built on top of this, not against it.
POLITICO Pro, FiscalNote, Quorum, Vuelio, Delphian, PoliVue — mature B2B political intelligence platforms in the US and EU.
None cover Ireland. All prove institutional buyers pay €500–2,000/month for structured political intelligence. DáilDex is the Irish equivalent, citizen-first.
Business model
The fastest revenue does not require a consumer app, a large user base, or a full product build. Structured political data has institutional buyers today. The consumer product compounds the value; it is not a prerequisite.
Email The Journal, Irish Examiner, and Waterford Live before the app launches. Offer structured TD data via API. €500–2K/month each. One yes pays VPS and LLM costs for a year. No app needed — just working data and a sample package.
Cold email UCD, TCD, and Maynooth politics departments. Annual access to structured TD voting and activity data for research. They are already building this in Excel, badly. Zero ongoing maintenance.
Non-dilutive funding designed exactly for scalable Irish startups with export potential. DáilDex qualifies cleanly. Application takes two weeks. No users required — just a credible plan and a working demo.
Before the app ships, a simple landing page charging €29.99/yr upfront for founding member Pro access. Irish political Twitter is small and engaged. 200 founding members proves willingness to pay before the iOS code is written.
lobbying.ie is already public. A thin AI monitoring layer on top, sold to law firms and public affairs consultancies at €199/month. Completely standalone from the main app — same underlying infrastructure.
Revenue sources
Institutions are the best early revenue. Politicians can pay for factual reports. Citizens anchor the trust and the product. The money is mainly B2B — organisations that need clean political intelligence at €200–2,000/month rather than citizens at €20–50/year.
They pay because structured data saves hours and surfaces stories their teams miss.
Only if the product stays factual and transparent. No reputation management, no attack research.
Most citizens will not pay €20–50/month for politics. B2B customers pay €200–2,000/month.
Pricing
The priority is becoming the place people check the public record — not squeezing early users. Data is the moat only when the dataset is trusted, source-backed, and used repeatedly.
TD profiles, vote records, PQs, debate summaries, party pages, national briefings, basic AI chat, and source links. Political transparency should not be paywalled. This is the trust layer.
Unlimited AI chat, advanced constituency alerts, saved searches, historical archive depth, digest emails, data exports, and promise tracking. Designed for journalists, activists, and engaged citizens.
API access, bulk exports, cross-TD analytics, custom topic alerts, webhooks, and research dashboards. Target: media companies, universities, NGOs, public affairs firms, and compliance teams.
Structured datasets, election intelligence packs, constituency reports, party discipline analysis, and sector monitors. Revenue without weakening the free civic layer.
Strategy
Media API — Academic licence — Lobbying monitor
Consumer DáilDex — Pro waitlist — Transparent sourcing
CouncilDex — Seanad/MEP layer — Election intelligence
ParliamentDex — White-label — Franchise model
Architecture
Official public records are the hard source of truth. Hermes agents run continuously over them — summarising, classifying, detecting stance changes, and building the cross-TD knowledge graph. Postgres stores every verified fact. The product surfaces it. The separation is intentional: credibility depends on proving exactly where each claim came from.
Official Record, Bills Office, Lobbying Register
160 TDs watched 24/7 — continuous ingestion and processing
Every vote, PQ, speech, stance stored with source links
Persistent AI memory per party — positions, discipline, history
Profiles, alerts, AI chat, institutional reports
AI infrastructure
The key insight investors miss: DáilDex agents do not run per user. Hermes monitors the Dáil once — continuously — and the output is streamed to every citizen, journalist, and API consumer simultaneously. Real-world agent costs run roughly 5-6x above raw per-token rates (multi-step workflows, tool calls, retries, context assembly) — but the total is still under €2K/year for all 160 TDs. Even handling 50+ concurrent users, the monitoring cost is fixed. The user base is unlimited.
Full week of monitoring one TD — votes, PQs, speeches, summaries.
Monitor all 160 TDs for an entire week. Total. Not per user.
Annual agent cost for the full Dáil. Whether we have 50 or 100,000 users.
Users who receive that output. Fixed cost, unlimited distribution.
30B parameter hybrid Mamba-2 + Transformer MoE with only 3.2B active parameters per token. 3.3x faster than comparable models. Open source under NVIDIA Open Model License. Powers all continuous TD monitoring — 160 agents running continuously. Agent costs are roughly 5-6x higher than base per-token rates due to multi-step agentic workflows, tool calls, and retry logic — still under €2K/year for the entire Dáil.
~€0.22 per TD per week. Entire Dáil: ~€35/week.
1.6 trillion parameter MoE with 49B active parameters, 33T training tokens, 1M context window. Released April 2026. Codeforces rating 3,206 — sits between GPT-5.2 and GPT-5.4 on reasoning. Available for deep analysis, institutional reports, and power-user AI chat.
Used on-demand. Users who want deeper analysis opt in.
Google's open 31B dense model released April 2026. 89% AIME, 84% GPQA. Excellent for structured summarisation, citation extraction, and multi-document synthesis. Runs at a fraction of frontier cost with near-frontier quality on political text tasks.
Best cost/quality ratio for summarisation workloads.
398B sparse MoE (13B active), 17T training tokens, 512K context. 94.7% on τ²-Bench. Purpose-built for multi-step agentic workflows with explicit chain-of-thought reasoning. Apache 2.0 licence. Used for complex cross-TD analysis and institutional research tasks requiring full audit trails.
Chain-of-thought reasoning for complex political analysis.
Product
Not a newspaper, not a party app, not a government portal, not an outrage feed. A neutral intelligence surface over public democratic records — free for citizens, valuable for institutions.
Find your TD, follow parties, track votes and PQs, compare records, receive alerts, and understand the week without political jargon. Works from any Eircode.
Monitor political groupings instead of running hundreds of isolated agents. Rooms retain context about positions, discipline, promises, topics, and recurring watchpoints.
Answers only after retrieval from DáilDex records. If the record is thin, the answer says so — no invented certainty, no hallucinated confidence.
Structured data feeds, webhooks, bulk exports, and cross-TD analytics for media companies, academics, NGOs, and public affairs teams who need the raw intelligence.
Defensibility
160 TDs, 24/7, with no manual effort. The moment something happens in the Dáil, it is processed, classified, and surfaced. No competitor or journalist team can replicate this at this cost.
The asset compounds as records, summaries, stances, promises, corrections, and source links accumulate. Each day the dataset is harder to replicate. Competitors start from zero.
Party rooms develop deep contextual memory about positions, spokespeople, recurring files, discipline risk, and issue history. That memory is the product — not just the raw data underneath it.
DáilDex describes patterns, it does not attack politicians. That credibility is hard to rebuild if lost. The moat is trust compounded by time, not just technology.
Expansion
Once the source-monitoring, room, retrieval, and product layers work for Dáil records, the same harness points at adjacent institutions and other parliaments. The build cost is amortised; the revenue surface grows linearly.
Councillors, council meetings, planning decisions, declarations, and local accountability where tooling is weakest. The best expansion after DáilDex.
lobbying.ie public records turned into a searchable, alertable risk and journalism layer. Clear buyer, clear pain, clear budget. Less dependent on consumer adoption.
Candidate records, constituency briefings, issue comparisons, and live election-period research tools. High demand, time-bound, high willingness to pay.
License the monitoring harness to media or civic groups in Westminster, Ottawa, Canberra, and Wellington. Same infrastructure, zero incremental build cost.
Complete the Irish political picture and bridge into EU-level monitoring. Obvious expansion of the same source graph and room architecture.
Find local operators in the UK, Canada, or Australia. Give them the full stack and brand licence, take 20% of revenue forever. Zero capital expenditure, infinite geographic scale.
Build status
Honest. The prototype is working; the production data pipeline is the next milestone.
The ask
Turn the prototype into Ireland's default political record. Hire the team. Get DáilDex on the street.
Two full-stack engineers and one growth lead. The product is built — now we need the team to ship v1, run agents at scale, and own the go-to-market.
Billboard campaigns at Leinster House and commuter corridors. Constituency-targeted local ads. University partnerships. Media API outreach to newsrooms. EI HPSU application.
Legal, compliance, 18-month operating runway, and contingency for early institutional pilot onboarding.
LLM inference, Supabase hosting, monitoring, and the Hermes agent compute layer for 160 continuous TD watchers.
Milestones to Series A
Why this is unusual
The first three revenue plays require no consumer traction. Media API, academic licensing, and EI grant funding are available before the iOS app ships. That is rare proof of concept for early-stage civic tech.
Political neutrality
DáilDex is a mirror, not a megaphone. We have no political position — we report what TDs say, how they vote, and whether those two things match. That principle is enforced at every layer of the stack.
Every model we run — Nemotron, Gemma, DeepSeek, Arcee — is open-source and hosted on Western infrastructure. No black-box API calls to closed models. Every weight is auditable.
We do not fine-tune models on political data. Fine-tuning creates unauditable bias baked into weights. Instead, we use carefully engineered system prompts that instruct the model to report facts and cite sources — nothing more.
Every system prompt that powers DáilDex agents will be published publicly. Anyone — journalists, academics, TDs themselves — can read exactly what the AI is told to do. Total transparency.
All inference runs on EU and US infrastructure. No data routes through jurisdictions with state-level AI censorship or filtering. Irish political data stays on trusted Western servers.
The founder
I have had the privilege of accessing frontier models before public release and providing direct feedback to the teams building them. DáilDex is not a first encounter with cutting-edge AI — it is the application of years spent inside the ecosystem, now directed at local civic infrastructure.
Selected to test and provide feedback on Kimi-K2.6 before public release. Worked directly with Moonshot AI on model behaviour, agentic capabilities, and real-world task performance.
Provided pre-release feedback on both GLM-5v and GLM-5.1 for Z.AI (Zhipu). Evaluated multimodal reasoning, instruction following, and edge-case handling across both model generations.
Community support member in the early days of OpenAI’s public Discord, helping developers and researchers navigate the API, prompt engineering, and model behaviour during the GPT-3 and early GPT-4 era.
I know AI well — the models, the inference economics, the agentic architectures, and the failure modes. DáilDex exists because I want to apply that knowledge locally, to benefit the people who actually live here. Not another Silicon Valley SaaS. A civic tool, built by someone who understands the technology deeply enough to make it serve a democracy.
Get in touch
DáilDex is a lean, opinionated team building the intelligence layer that Irish democracy is missing. We are looking for aligned capital, data partners, and institutional pilots — not a dozen simultaneous conversations. If the thesis resonates, reach out directly.