status: accepting_clients
gofreddy.ai AI marketing agency

Marketing,
engineered.

An agency that runs AI marketing systems with humans in the loop. We run multiple AI sessions per task, engineers and humans review every output, and we iterate until we have the best actions to recommend to you.

Not autonomous robots. AI agents that draft, judges that score, engineers that review, iteration that improves. Everything visible on your portal.

Book audit See how it works
$1K
Audit from
1w
Turnaround
6
Client cap
Live from a client portal
freddy.live acme.com · session_42
09:42:13 keyword_miner.start { query:"acme distribution surface", depth:3 }
09:42:14 ✓ 247 long-tail keywords surfaced · cost $0.04
09:42:18 competitor_scout.scan { rivals:12, since:"7d" }
09:42:21 ✓ 3 new ad creatives detected · 1 priority match
09:42:24 headline_writer.draft { angles:5, model:"opus-4.7" }
09:42:37 content_judge.score { drafts:5, axes:9 }
09:42:38 ⚑ top 3 drafts queued for engineer review
09:42:51 human jan.review → approved draft_3, edited draft_1
09:42:51 publisher.queue { platform:"x.com", at:"14:00 CET" }
09:42:51 ✓ scheduled · session cost $0.31
09:42:52 $
The stack
§ 01 · What I build

What we ship.
AI sessions + human review.

10 production engines · 4 surfaces · same voice persona + audit trail across all of them

One loop, five steps. Our AI agents practice on themselves in the background. When we start your project, we pick the version doing best in your kind of domain. The agents do the actual marketing work. An engineer plus a human reviews every output before it ships. What humans accept (or edit) feeds back into step one — so the agents keep learning from real work, not just from tests. (Inspired by Karpathy autoresearch and adjacent self-improving systems.)

01 · Text · 3 engines

Content factories.

Long-form articles + blog SEO. Short-form posts on X + LinkedIn. Email + newsletter copy. Same voice persona consumed by every channel — no rewriting per surface.

Outputs: daily-to-weekly cadence · ranked drafts with reviewer notes · what shipped + why.
article_engine · x_engine · linkedin_engine
02 · Visual + video · 3 engines

Creative production.

Static visuals (IG carousels, LinkedIn document carousels, hero images, ad statics). Short-form video for TikTok, Reels, Stories. Paid ad creative for Meta + LinkedIn.

Outputs: AI-generated previews · scene plans · voice scripts · review-ready before ship.
image_engine · storyboard · ad_engine
03 · Signals · 3 engines

Intelligence + monitoring.

Continuous scan of competitors, brand mentions, AI-search readiness, share-of-voice. Ad libraries, news, social, Reddit, podcasts, reviews, search. Ranked signals — not a dashboard of noise.

Outputs: weekly digest · real-time competitor alerts · monthly market briefs.
competitive · monitoring · geo
04 · Visibility · 1 engine + portal

Diagnostics + reporting.

149-lens marketing audit (the $1K–$3K phase 01). Live portal: agent transcripts, cost ledger, reviewer events, audit trail. Compliance gating where your industry requires it.

Outputs: HTML reports · raw JSONL · cost-per-output ledger · full audit trail.
marketing_audit · client_portal

Custom pipeline not on this list? Tell me what work needs doing. We build new lanes when clients ask — quote and timeline scale with scope. The audit decides which standard ones you need; usually a mix.

§ 02 · The work

Two phases.
Each stands alone.

Fixed scope per phase. Fixed price per phase. You exit after either one and the work holds up. No 12-month retainers, no escalation ladders.

Phase 01
1 week

Audit

The diagnostic
$1K
– $3K
Refund if I don't surface 3 specific gaps

I map your distribution surface. Where customers find you. Where competitors beat you. Where the cheapest distance between you and more revenue is.

  • + 149 specific lenses across 9 measured dimensions
  • + 1000+ keywords · competitor ads · share-of-voice math
  • + Social + news + PR landscape · content quality scan
  • + Technical SEO + AI-search readiness
  • + Every finding sourced, severity-ranked, pipeline-tagged
  • + HTML report (memo-style) + raw JSONL attached
Not included: brand strategy · PR · paid ad management
Start here →
Phase 02
Ongoing

Operate

Build & run
$2K
– $10K/mo
Scope set by audit · smaller retainers possible

I build the engines your audit calls for and run them. Same outputs a marketing team would ship — content, publishing, intelligence, reporting — delivered week to week without you managing it.

  • + Pipelines built to your audit's spec
  • + Weekly optimization · 1–2 day max turnaround
  • + Live agent transcripts + telemetry on your portal
  • + Live HTML reports · quarterly strategy review
  • + Optional export: take an agent snapshot to self-host (no ongoing updates)
Default: Freddy runs the pipelines on our infrastructure · you watch the work on a portal. Optional add-on: point-in-time export of agents + runbooks for self-hosting — exports freeze on handoff and stop receiving improvements.
Requires audit →
§ 03 · The timeline

From email to running system
in about three weeks.

No discovery decks, no scoping calls, no Statements of Work. You send a domain and a paragraph. I send back when the audit starts.

Day 0

Brief.

Domain + one paragraph in

You email me your URL and one paragraph about your setup. I reply same-day with a scope estimate and the next available audit slot. No call required to get started.

Week 1

Audit.

149-lens diagnostic · 9 dimensions · every finding sourced

One week of structured research across nine measured dimensions of your distribution surface — organic search, paid + earned ads, social channels, news + PR mentions, competitor activity, share-of-voice math, content quality on your owned surfaces, technical SEO foundations, AI-search readiness. 149 specific lenses behind them, applied at depth — not a checklist scan. Every finding is sourced, severity-ranked, and tagged with a recommended pipeline (or an explicit build-or-pass call). HTML report reads like a memo; raw JSONL attached for your engineers.

Week 2–3

Build.

Built + run on Freddy infrastructure · agency model

I build the pipelines your audit calls for and run them on Freddy infrastructure. No integration into your stack — we're the agency operating the system; you get the outputs week to week and a portal that streams the work in real time (agent transcripts, costs, decisions, drafts queued for human review). Optional add-on: export the agents + runbooks to self-host on your own system. Exports are point-in-time snapshots — once they leave Freddy, they stop receiving updates.

Portal live · export optional
Week 3+

Run.

Live operation, weekly cadence

Pipelines online. Weekly optimization passes. Monthly performance reports tied back to specific bets. Quarterly strategy review. Systems self-improve under oversight — pipelines learn from each run and the next one starts smarter.

§ 04 · Structural differences

You've done this with an agency.
It doesn't have to be that.

Dimension Traditional agency Boutique AI studio Freddy
Headcount on your work 5–15 humans, account managers, slides 2–5 humans, founder + small team 1–2 engineers + multiple AI sessions per task
Monthly spend $15K – $60K retainer $8K – $25K retainer $2K – $10K / month
Iteration speed Weeks per change (approval cycles) Days per change 1–2 day max for quality work
Reporting Curated slides, monthly Branded dashboard you log into Live HTML reports · designed to read · raw JSONL attached
Visibility into the work Status emails, scheduled check-ins Monthly digest, occasional ad-hoc updates Live agent transcripts · streaming telemetry · full audit trail
What you actually get PowerPoints. Dashboards. Reports. Their codebase. You license access. Audit deliverables · live portal of the work · agents run on Freddy · snapshot export optional
When we part ways Marketing stops Licenses revert, dashboard expires Engagement ends · exported snapshots stay with you if you opted in (frozen at handoff)

Swipe table to compare →

§ 05 · Who runs this

A message from the founder.

Jan Noszczyk. Founder + builder. Built the AI systems that do the work of a marketing department — content, publishing, competitive intelligence, monitoring — first shipped as Freddy SaaS to paying users, now the same engine runs behind per-engagement builds.

Freddy the agency is the same engine, priced per engagement instead of per seat. AI marketing systems with humans in the loop — engineers and humans review every output the agents produce, and we iterate until we have the best actions to recommend.

Six is the maximum number of operate clients I take at once. Math: a week of attention per client per month, two weeks for new-client ramp, two weeks for slack. If I'm full, you go on the list.

noszczykai@gmail.com github.com/jryszardnoszczyk/gofreddy → LinkedIn →
Shipped
0
Source files
Integrations
0
Data providers
Models
0
Frontier LLMs
Python
3.13
Pinned
§ 06 · Known objections

Answered in advance.

01   what does freddy actually do? ›
Audit your distribution surface, then build and run the AI systems your marketing team would build — content pipelines, publishing, monitoring, competitive intelligence. Engineering work, not campaign work. Engineers + multiple AI sessions per task, humans review every output before it ships, results delivered on a portal you can read.
02   why not just hire an agency? ›
You can. Freddy is one option among many. The bet here is that most marketing work in 2026 is engineering — building and running AI systems with humans reviewing every output — and an engineer-led agency delivers it with less overhead and faster turnaround than a traditional 5–15 person team. If your needs are heavier on brand strategy, live events, or buyer interviews, a traditional agency might fit better.
03   can one person really replace a marketing team? ›
Depends on what the team does. Content research, competitive monitoring, publishing, reporting — yes, those. Buyer interviews, brand strategy, live events — no. The audit tells you which kind of team you have.
04   why so cheap? what's the catch? ›
No catch. No 15 salaries, no rented office, no sales team. You pay for output. Your ceiling is my calendar — which is why I cap at six operate clients.
05   bus factor — what if you disappear? ›
Depends on what you took. Audit-only: you keep the report and the raw data, you take them to whichever agency comes next. Operate as standard agency: I'd arrange continuity at handoff. Operate with an export taken: you have a point-in-time snapshot of the pipelines, runbooks, and keys — you can self-host or hand them to another engineer. Snapshots freeze at handoff date and stop receiving improvements, so most clients let us keep running the live versions; the export is there in case you need clean separation.
06   isn't this just chatgpt with extra steps? ›
ChatGPT is an ingredient. Freddy is the kitchen. Real distribution systems need data ingress, scoring, routing, cost control, audit trails, error handling, secret management, and someone on call when a provider changes their API. A chatbot has none of that.
07   what if the audit finds nothing? ›
If the audit doesn't name at least three specific distribution gaps and a plan to close them, you get a refund. Haven't had to issue one.
08   do I have to continue after the audit? ›
No. The audit is complete on its own. You get the report, the raw data, and full ownership. Operate is a separate decision, made with more information once the audit is in.
Ready 2 slots

Start with
an audit.

One week. From $1,000. Refunded if it doesn't surface three specific distribution gaps and a plan to close them. Email me your domain and one paragraph about your setup.

No campaigns· No calendars· No decks· No account managers

Book audit →

Reply within 24 hours · noszczykai@gmail.com