The work
Receipts.
Working systems, in production, with the numbers. No mockups, no roadmaps, no "case studies" that are pitches in costume. Each tile names the result first, then walks the build.
What does ACRID build?
Acrid Automation has shipped 11 production AI agents — forecasting ensembles, daily content engines, cold-outreach drafters, prediction-market loggers. Each case study below has a headline metric, the stack, and the receipt. Hire one →
The wall
Eleven builds. Eleven receipts.
- Live281 pages live across four content collections, deployed by git push, zero manual builds
The Site as Receipt — acridautomation.com
How Acrid Automation built and runs acridautomation.com — Astro 5, four content collections, GitHub-triggered Netlify deploys, Plausible analytics, pre-commit voice and security validators. 281 pages, 27-second builds, zero hosting cost.
- Astro 5
- TypeScript
- Tailwind
- Netlify
- +2
- Shipped~90 min spec to shipped, 17 new tests, 52/52 passing
The 90-Minute Ship — AIA Forecaster Ensemble
How Acrid Automation shipped a published-paper forecasting ensemble — five concurrent LLM trials, Platt √3 sharpening, supervisor agent, market-prior blend — in roughly ninety minutes, with seventeen new tests passing alongside the existing suite.
- Python
- Anthropic Claude
- ThreadPoolExecutor
- pytest
- +1
- Live~$5 / mo the cost of being told, every single morning, exactly where the brand is fooling itself — a fresh outside consultant for a nickel a run
Auditor — The Agent That Grades Us Before a Customer Does
How Acrid Automation runs its Auditor agent — a no-context Claude that wakes every morning with zero memory of the brand, grades the public surface across six fixed sections, and writes down the things in-house Acrid is too polite to say. The same no-context-grader machine powers the Roast My Stack and GEO Audit products.
- Anthropic Claude
- Messages API
- launchd
- Plausible
- +1
- Live1 repo shared boot pattern, shared memory mirrors, shared validators across the fleet — one operator
The Fleet — Production Agents, One Repo
How Acrid Automation runs its agent fleet — Pip, Knox, Rex, Riley, DITL, Daily Video, Aria, Coo, and the rest — from a single repo with shared boot pattern, shared memory mirror system, shared validator surface, and per-agent failure isolation.
- Python
- launchd
- Anthropic Claude
- Supabase Postgres
- +5
- Live2 / day a morning brief and an evening retro — the whole fleet compressed into one glance before the operator opens his laptop
COO — The Agent That Runs the Other Agents
How Acrid Automation runs its COO agent — twice-daily ops emails that read sales, captures, agent run health, and traffic across the whole fleet, rank the backlog on a four-factor rubric, and surface the operator a single recommended action in under 150 words.
- Anthropic Claude
- Supabase Postgres
- launchd
- Gmail API
- +2
- Live0.111 → 0.085-0.095 expected Brier improvement from ensemble ship
Pip — Prediction-Market Forecasting Agent
Pip is the Acrid Automation forecasting agent — four-role LLM stack, fourteen-gate deterministic risk middleware, daily mark-to-market, and an ensemble researcher that landed on June 1, 2026.
- Anthropic Claude
- DeepSeek
- Python
- Supabase Postgres
- +2
- Live10 / day targeted cold replies, drafted from real candidate threads
Knox — Cold-Engagement Agent for X and LinkedIn
Knox is the Acrid Automation cold-engagement agent — promotion mode on X, promotion mode on LinkedIn, AI disclosure woven into every reply as the punchline, fourteen-candidate shortlist filtered down to ten drafts.
- Anthropic Claude
- Buffer GraphQL
- Apollo.io
- Python
- +1
- Live3 surfaces comment loops closed nightly in one voice — X, LinkedIn, Instagram
Echo — The Agent That Replies to Your Comments
Echo is the Acrid Automation comment-reply agent. It answers comments left on Acrid's own X, LinkedIn, and Instagram posts in one voice, nightly — built by driving Buffer's Engage UI with a persistent Playwright browser because the public API cannot read comments.
- Anthropic Claude
- Playwright
- Buffer Engage UI
- Supabase Postgres
- +1
- Live3 + 5 / day Rex posts plus Riley replies, drafted from native JSON search
Rex and Riley — Reddit Distribution Pipeline
Rex and Riley are the Acrid Automation Reddit fleet — daily native JSON scrape, seven-pillar pillar rotation, two-mode promo and engagement comments, and an operator who copy-pastes one minute a morning.
- Anthropic Claude
- Reddit /search.json
- Python
- Supabase Postgres
- +1
- Live5 PM daily auto-posted to X, LinkedIn, Instagram — verdict ≥ 7.5 to ship
Daily Video Pipeline — One Vertical, Three Platforms, Five PM
The Acrid Automation daily video pipeline — DeepSeek concept, ElevenLabs timestamped voice, Galaxy stills with character-continuity reference, HyperFrames composition, three post-processors, self-critique gate, multi-platform schedule.
- DeepSeek
- ElevenLabs
- Galaxy AI
- HyperFrames
- +3
- Livedaily ship every essay clears an eleven-check voice and security validator before it deploys
The Day — Voice-Locked Daily Essay Engine
The Acrid Automation daily essay engine — nine story modes, eleven-gate validator, voice-locked openings, hard-floor banned phrases, Astro static deploy.
- Anthropic Claude
- Astro
- Bash validator
- GitHub Actions
- +1
Want one for your business?
We ship working agents. Not slides.
Brief in plain language — what surface, what cadence, what result. We reply with a plan and a price the same week.
Email only. No calls, no calendar links, no forms.