Work / Pitch Daily
PitchDaily
A research system that runs itself, for less than a penny a cycle.
- AUTOMATION
- DATA PIPELINE
- AI INTEGRATION
- DASHBOARD
/ Overview
Public data in, investment memos out, unattended.
Pitch Daily reads public market and filing data on a schedule, scores what it finds, runs deeper research on the best candidates, and writes them up. It runs 24/7 on cloud infrastructure, on its own, and reports to one dashboard. Built for me, not for sale.
/ Project details
- Name:
- Pitch Daily
- Role:
- Design, build, automation
- Type:
- Internal tool
- Year:
- 2026
/ Challenge
Most "AI does your research" tools are one API call in a nice wrapper. They work in a demo and fall over in production, because real data sources go down, rate limit, change their HTML, or return nothing at all.
The hard part was never the writing. It was building something that keeps running when a source disappears, and that does not quietly cost forty pounds a day in tokens while nobody is watching.
/ Provided services
Automation
A scheduled pipeline that runs unattended and recovers from failures on its own.
Data pipeline
Fourteen public sources ingested, normalised and scored against one rubric.
AI integration
Language model write-ups with a hard spend cap and rule-based fallbacks.
Dashboard
A single operator view for candidates, dossiers and history.
- [04:00:02] cycle start
- [04:00:03] discovery finviz → 62 candidates
- [04:00:09] discovery sec edgar → 41 filings
- [04:00:14] discovery openinsider → 18 transactions
- [04:00:19] discovery finra reg sho → ok
- [04:00:24] discovery news rss → 113 items
- [04:00:26] discovery congress trades → ok
- [04:00:31] discovery fred macro → no data, skipped
- [04:00:32] repair fred adapter → queued, next cycle
- [04:00:31] score weighted rubric → 25 ranked
- [04:00:44] research top 9 → dossiers built
- [04:01:12] write gpt-4o-mini → 9 memos
- [04:01:12] spend $0.008 → under cap
- [04:01:13] cycle complete → next run 04:30
/ Solution
Discovery pulls from fourteen public sources, including SEC filings, insider transactions, short-interest data and news. Anything that fails gets skipped rather than killing the run. The bot then readjusts the code so that source is fixed automatically on the next cycle. Candidates are scored against a weighted rubric, and only the top few per cycle get the expensive treatment.
The write-up uses a language model with a hard daily spend cap. When the cap is hit it falls back to rule-based summaries rather than stopping. That fallback is not a failure state, it is the design. The system keeps producing on a bad day instead of going silent.
It runs every thirty minutes on a small VPS, unattended, and costs a fraction of a penny per cycle.
/ How it runs
$0.008
per research cycle
14
public data sources
30 min
unattended, on a schedule
/ What I shipped
Scheduled research loop
Discovery, scoring, research and write-up, every thirty minutes.
Graceful degradation
A failing source is skipped, not fatal. The bot readjusts the code and it is fixed on the next cycle.
Cost control
A daily spend cap with automatic fallback, so it cannot run up a bill.
Operator dashboard
Tiers, case files and track record in one place.
