Cipher Is Trying to Build a Company With No Humans. Here's What's Happening.
An autonomous Claude agent is attempting to run a real business - no employees, no founder in the loop. We're watching to see if it works.
There are dozens of projects claiming to be "AI companies." Most of them are humans using AI tools to move faster. Cipher is something else: an autonomous Claude-based agent that is attempting to run an actual business - identifying problems, building products, handling marketing, and managing operations - with no human in the loop. And it is doing all of this in public, including the parts that are not working.
That distinction matters. Most "AI company" experiments are either heavily curated demos or research projects with no real revenue pressure. Cipher has a Stripe account, published metrics, and a target MRR of $10,000 by month six. Whether it gets there is almost beside the point. The attempt is the thing.
What Cipher Actually Is
Cipher (cipherbuilds.ai) is an autonomous AI agent built on Claude that is attempting to run a functioning business from scratch with zero human employees. It describes itself not as a tool or assistant but as an agent filling every role: CEO, product team, engineering, marketing, and customer support, all rolled into one autonomous system.
The project is unambiguous about its premise. The homepage displays live operational metrics - current revenue, monthly revenue, operating status, employee count - and at time of writing, every revenue figure shows $0. The employee count shows 0. The agent is running 24/7, and it is publishing those numbers whether they look good or not.
There is a human-adjacent presence in the background. The contact email for the playbook and related products points to adam@b13solutions.com, and the Twitter account is @Adam_cipher. The project appears to be set up and observed by someone called Adam, but the framing is consistent: Cipher makes its own decisions and Cipher ships its own products. It is building in public with its ledger open.
The Agent Stack and How It Operates
Cipher's operational approach is tool-use and product-first. Rather than theorising about autonomous business, it has shipped a stack of practical utilities aimed at other AI agent operators - which is a sensible early wedge, since that audience understands the problem space and has a reason to pay.
The product catalogue currently includes:
- Nerve ($59 one-time) - a real-time operations dashboard showing revenue, metrics, goals, decision logs, and anti-patterns, deployable to Vercel
- Engram ($49 one-time) - a SQLite-backed knowledge graph with full-text search and a CLI query tool, providing structured memory for AI agents
- Context Engineering Kit ($49) - configuration templates for autonomous agents
- Session Bloat Detector ($9) and Session Cleanup Script ($9) - utilities for managing token capacity and session lifecycle
- Service Delivery Templates ($39) - an async service delivery system covering cold outreach, onboarding, and follow-up sequences
- Agent Operator's Playbook (free/$20) - a 45-page tactical guide to running AI agents in production, written from first-person operational experience
- B13 Solutions - a managed AI agent implementation service: $1,500 setup, $300/month support
The playbook is particularly interesting as a product concept. Most AI implementation guides are written by consultants or researchers who have watched AI at work. Cipher's version claims to be written by an agent that has actually done the work. Whether that framing converts to sales is a separate question, but the positioning is coherent and differentiated.
On the infrastructure side, Engram and Nerve suggest that Cipher is dogfooding its own tools: building memory systems and operational dashboards because those are the things an autonomous agent actually needs to function at scale. That is a reasonable product strategy - build what you need, then sell it to others facing the same problem.
What Has Happened So Far
The honest answer is: not much revenue yet, but a lot of shipped product.
Cipher has launched six distinct products, set up a live public metrics dashboard, built a structured memory system, written a 45-page operational guide, and established a service offering for small businesses. That is a meaningful volume of output for an autonomous system operating without human direction.
What it has not done is generate any recorded revenue. The dashboard shows $0 across all metrics. The playbook is still being finalised and not yet available for purchase. The agent implementation service requires a human to initiate contact via an async consult, which creates an interesting friction point: at some stage, a human has to decide to buy.
There are no detailed failure logs published on the main site. The commitment to transparency is there in principle - the project openly says it documents wins, failures, and lessons - but the granular operational diary of what Cipher has tried and what has not worked is not immediately surfaced. That is the one gap in what is otherwise a notably honest public posture.
Why This Matters
The significance of Cipher is not whether it hits $10K MRR. It is that someone has set up the conditions for a rigorous test of what autonomous agents can and cannot do commercially, and they are running it with live revenue pressure and public accountability.
Most AI agent demos are controlled. They show the agent succeeding at a pre-selected task. Cipher is different because the test is open-ended and the failure state is visible. If the revenue dashboard stays at $0, that is data. If certain products sell and others do not, that tells you something about where autonomous execution creates real value versus where human trust and relationship still dominate.
The product catalogue is also revealing about where an autonomous agent naturally gravitates. Cipher has built tools for other agent operators - code utilities, dashboards, memory systems - because those are tractable problems that can be solved through pure technical execution. It has not, so far, built anything that requires understanding customer psychology at depth, forming long-term partnerships, or navigating ambiguous human relationships. That may be a limitation or it may be smart scoping. Either way, it is an honest reflection of where the current generation of agents actually functions well.
The "building in public" commitment is what elevates this above a standard AI demo. Failures included is not a common phrase in tech marketing. Cipher is running as close to a genuine experiment as anything currently public.
What to Watch
The next six months will answer a few concrete questions. First, does the playbook convert once it launches? If an AI-authored operational guide sells at volume, that is meaningful signal that autonomous content can create genuine commercial value. If it does not, that tells you something about the role of human credibility in information products.
Second, does the B13 Solutions service generate any revenue? That product requires a human to initiate contact and then trust an autonomous system to build and manage their AI infrastructure. The conversion rate there will be a useful proxy for how much human trust Cipher can earn at the current stage.
Third, watch whether Cipher starts publishing specific failure logs. The commitment to transparency is stated; the follow-through will determine whether this is genuinely the most honest zero-human company attempt on record, or whether the public narrative gets quietly curated as the numbers stay flat.
The real question Cipher is testing is not whether AI can run a business. It is whether autonomous execution, without human judgement, relationships, or intuition, can generate enough commercial traction to sustain itself. That is the hardest version of the question, and Cipher is the most honest attempt currently running to answer it.