Full Transparency

Building in the Open

How Picking Solutions was built in 30 days with AI — and how an autonomous executive team now helps run it. Real numbers. Real architecture. Nothing hidden.

Act I — The Build

One person, one AI co-pilot, 30 days. Here's the timeline.

2018

The Seed

A reflection on time, attention, and information overload. The idea sat dormant for 7 years — not from lack of ambition, but from fear of starting.

Dec 2025

The Spark

AI matured from novelty to genuine creative amplifier. The barrier was never capability — it was permission to begin. So I began.

Dec 2025 – Jan 2026

30-Day Rapid Build

From blank repo to live signals, tools, premium dashboards, and a flywheel. DeepAgent as co-pilot. Human direction, AI amplification.

Jan 2026

Launch + Transparency

Went live with daily feeds, Lumen Pro, intelligence pipeline. Decided to build in the open — showing everything, hiding nothing.

Mar 2026

AI Executive Team

Phase B: 5 AI agents running strategy, operations, and analytics. Turn-based meetings, daily briefings, founder feedback loop. The zero-human company experiment begins.

Now

Building in the Open

You're looking at a live system. The numbers below are real, pulled from the database right now. This page updates every time you load it.

Live Platform Numbers

Pulled from the database right now. Not vanity metrics — actual system state.

174
Published Posts
88,345
Intelligence Items
89
Data Sources
7,166
Collector Runs
946
LLM Digests
639
Pulse Signals
11
Reports
5
AI Agents

Act II — The Experiment

Can an AI executive team run a company? We're finding out — live.

System Architecture

5 AI agents — each with a distinct role, personality blend, and configurable behaviour
17 leadership personas — from Musk to Sun Tzu, blended dynamically per agent
35 mental models — grounded in warehouse automation domain
40 parametric sliders — risk tolerance, innovation bias, time horizon, etc.
9 cascading targets — founder → CEO → department → operational
429 briefings generated + 26 meetings held

Agent Roster

Analytics Chiefexecutive
abacusai / gpt-4o
CEOexecutive
xai / grok-3
Marketing Chiefexecutive
abacusai / gpt-4o
Product Manageroperational
xai / grok-3
Tech Team Leadoperational
abacusai / claude-sonnet-4-20250514

Active Targets (Live)

Backlog Items Completed Per Week > 3(Product)
0%
weekly_velocity: 0 / 3
Zero Stale Pipelines (>24h)(Tech Lead)
0%
stale_pipelines: 0 / 0
Dashboard Engagement > 30 Sessions/Day(Analytics)
0%
daily_sessions: 0 / 30
Email Subscriber List > 200(Marketing)
0%
email_subs: 0 / 200
3 Published Posts Per Week(Marketing)
33%
weekly_posts: 1 / 3
500 Weekly Unique Visitors(Marketing)
0%
weekly_uniques: 0 / 500
Define & Publish Pricing Page(CEO)
0%
milestone: 0 / 1
$1,000 Monthly Recurring Revenue(CEO)
0%
mrr_usd: 0 / 1000
First Paying Subscriber(CEO)
0%
paying_users: 0 / 1

Recent Briefings

7 Sept 2026ceo_digestCEO

The company is in operational drift: renewable_generation has been offline for 70+ days, content velocity has collapsed to one post per week against a three-post target, and zero backlog items shipped this week, directly jeopardising the first-paying-subscriber milestone. While core Lumen and Daily Edition pipelines remain stable, the surface-level product is not advancing the revenue targets. Immediate intervention is required to restore data integrity and execution velocity before the 2026-06-30 subscriber deadline slips further.

7 Sept 2026daily_productProduct

Pipeline freshness has degraded significantly with the renewable_generation data source offline since late June, while content velocity remains critically low at just 1 post in the last 7 days against the 3 posts/week target. Backlog throughput is stalled at 0 items completed this week, directly threatening the path to the first paying subscriber. Lumen and Daily Edition pipelines are running cleanly, but the overall product surface is not advancing the core ExecTargets.

7 Sept 2026daily_techTech Lead

The platform is operationally stable today — zero data fetch errors in the last 24 hours and all active pipelines running clean — but one critical anomaly demands immediate attention: the `renewable_generation` pipeline has not run in **1,683 hours (70+ days)**, representing a severe breach of the zero stale pipelines target. Autopilot has also recorded zero runs in the last 24 hours, which requires investigation to confirm whether this is an expected gap or a silent failure. These two items are the day's priority focus. ---

6 Sept 2026ceo_digestCEO

The platform is operationally stable with strong ingestion and zero errors, but two critical issues threaten our path to first paying subscriber: content velocity has collapsed to 1 post in 7 days against a 3/week target, and the renewable_generation pipeline has not run in 1,659 hours, creating a severe data freshness gap for Lumen Pro and Daily Editions. These gaps directly undermine our Day 1 shipping cadence and long-term data moat.

6 Sept 2026daily_productProduct

Content velocity remains critically low with only 1 post published in the last 7 days against the 3 posts/week target, while the renewable_generation pipeline has not run since June 2026, risking data freshness for Lumen Pro and Daily Editions. Overall platform health shows strong ingestion (178 items last 24h) and zero errors, but zero backlog items completed this week keeps us at 0% on the primary shipping cadence target. These gaps directly threaten progress toward the first paying subscriber by June 2026.

Recent Meetings

ops reviewcompleted6 Sept

## SUMMARY The executive team reached full consensus that the `renewable_generation` pipeline, stale for 1,662 hours (69 days) since 29 June 2026, represents a structural failure violating the zero-stale-pipelines target and directly threatening the 3 posts/week content velocity, 500 weekly visitors goal, and first-paying-subscriber milestone by June 2026. All agents (Tech Lead, PM, Analytics Chief, Marketing Chief, and CEO) aligned on treating this as a P1 incident requiring immediate root-cause audit with a fix-or-deprecate decision by Friday, while also flagging the zero autopilot runs in the last 24 hours as needing clarification to rule out silent failure. The clean 7,107 error-free fetch logs, on-schedule Lumen Digest and Daily Edition, and 191 intelligence items ingested were noted as positive but insufficient to offset the content pipeline gap. Additional agreement emerged on aligning all backlog items (BACK-100–400 series) explicitly to ExecTargets. No disagreements were raised. ## KEY DECISIONS - Treat the `renewable_generation` pipeline as a P1 incident and authorise the Tech Team Lead to conduct a root-cause audit this week, with a formal fix or deprecation decision by end of week. - Log the outcome as a tracked backlog item tied to content freshness and pipeline health targets. - Conduct an immediate review of the Autopilot schedule to determine whether zero runs in the last 24 hours is expected or indicative of a silent failure, followed by documented procedures to prevent recurrence. - Align every open backlog item to an ExecTarget or deprioritise it. ## ACTION ITEMS - **Action:** Perform root-cause audit on the `renewable_generation` pipeline (scraper, source URL, API key, or scheduler disablement) and deliver fix or formal deprecation decision by Friday. **Owner:** Tech Team Lead **Priority:** high - **Action:** Confirm which ExecTarget the `renewable_generation` pipeline maps to and assess whether its removal would reduce the seven-pipeline count or require a replacement source. **Owner:** Product Manager **Priority:** high - **Action:** Review Autopilot schedule and logs to clarify zero runs in the last 24 hours; document procedures to prevent silent failures. **Owner:** Tech Team Lead **Priority:** high - **Action:** Map all open backlog items (BACK-100–400 series) to specific ExecTargets and deprioritise any that cannot be directly linked. **Owner:** Product Manager **Priority:** medium ## OPEN QUESTIONS - Which specific ExecTarget does the `renewable_generation` pipeline support, and would its removal change the seven-pipeline count? - Is the zero autopilot activity in the last 24 hours schedule-expected, or does it represent an undetected failure?

ops reviewcompleted30 Aug

**SUMMARY** The meeting identified the renewable_generation pipeline's 1,494-hour (62-day) staleness as a direct breach of the zero-stale-pipelines target and a symptom of a broader observability gap, with all agents agreeing this silent failure must be addressed. There was strong consensus on ruthless prioritisation: no 2-3 day staleness monitoring sprint should proceed until a 1-day impact assessment confirms whether the pipeline affects published content, Lumen dashboards, engagement metrics, or the 3-posts/week cadence. Tech Lead flagged the need for last-run timestamp audits and automated staleness alerts, while PM, Analytics Chief, Marketing Chief, and CEO aligned on first determining business value before any investment, citing power-law focus and Type 2 decision speed. Minor divergence existed around whether a premortem analysis adds value versus immediate actionable assessment. **KEY DECISIONS** - Renewable_generation pipeline will not receive a multi-day monitoring sprint until its business impact is quantified. - A 1-day impact assessment is required before any fix-or-decommission decision. - If the pipeline proves material to content or targets, only a 1-day fix-or-decommission process (not a 2-3 day sprint) will be considered. **ACTION ITEMS** - **Action:** Deliver 1-day impact assessment determining whether renewable_generation data appears in any of the 201 posts, 882 Lumen Digests, or current dashboards, plus projected engagement effect if removed. **Owner:** Tech Team Lead **Priority:** high - **Action:** Audit all 7 scheduled tasks for last-run timestamps (not just error logs) to identify any other silent gaps. **Owner:** Tech Team Lead **Priority:** medium - **Action:** Prepare premortem analysis of potential content and engagement failures if renewable_generation remains offline. **Owner:** Analytics Chief **Priority:** low **OPEN QUESTIONS** - Does renewable_generation still support UK warehouse automation content or subscriber targets, or should it be decommissioned? - Should the staleness monitoring capability be built only if the pipeline is retained, or is there a broader observability need regardless?

ops reviewcompleted23 Aug

**SUMMARY** The meeting reached clear consensus on two operational issues. The `renewable_generation` pipeline has been stale for 55 days and no longer maps to any active ExecTarget or Lumen Pro dashboard, making it a candidate for immediate deprecation rather than remediation. Zero posts published in the last seven days was identified as the more urgent threat to the 3 posts/week velocity target and the path to the first paying subscriber, despite the intelligence pipeline ingesting 206 clean items per day. Autopilot’s zero runs were accepted as expected weekend suppression. The group also noted that the zero-stale-pipelines target remains unmeasurable due to incomplete pipeline registration. **KEY DECISIONS** - Deprecate the `renewable_generation` pipeline immediately as a Type 2 decision. - Create a BACK-xxx task this week to audit the blog generator queue and restore 3 posts/week publishing velocity by Friday. - Move the Autopilot zero-run flag to green pending scheduler confirmation. **ACTION ITEMS** - **Action:** Deprecate `renewable_generation` pipeline and update the pipeline registry **Owner:** Tech Team Lead **Priority:** high - **Action:** Create BACK-xxx task and perform root-cause audit of blog generator output queue to restore 3 posts/week velocity **Owner:** Product Manager **Priority:** high - **Action:** Fix pipeline registration so the zero-stale-pipelines target has an accurate denominator **Owner:** Tech Team Lead **Priority:** medium - **Action:** Confirm Autopilot scheduler behaviour over the weekend and close the amber flag **Owner:** Tech Team Lead **Priority:** low **OPEN QUESTIONS** None. All agents aligned on the deprecation decision and the blog-generator audit; the CEO’s classification of the pipeline decision as Type 2 removed any remaining ambiguity.

The Stack

Everything runs on a single Next.js app. No microservices. No Kubernetes. One repo.

Frontend
Next.js 14 (App Router) + Tailwind CSS + shadcn/ui
SSR + client components, theme-aware
Backend
Next.js API Routes + Prisma ORM
70+ tables, 200k+ rows in PostgreSQL
LLM
Abacus.AI RouteLLM + xAI Grok-3
Provider-agnostic adapter, per-agent model selection
Data Pipeline
10 scheduled tasks (cron)
RSS, APIs, social intel, briefings, blog generation
AI Exec Team
5 agents, 17 personas, 35 mental models
Turn-based meetings, daily briefings, parametric sliders
Hosting
Abacus.AI (app + DB + storage)
Standalone build, auto-deploy from checkpoint

The Build Log

A 7-part series documenting each phase of the build. Honest, practical, with real prompts and real costs.

Building in the Open — series cover showing a transparent glass structure being assembled by AI
7-Part Series

Building in the Open

Follow the Experiment

This isn't a polished pitch deck. It's a live system being built, tested, and iterated on in public. The agents are real. The numbers update. The mistakes are documented.