Now live — $199K collected, $0 raised

The AGI Revenue & Capital Layer
for the 400 Million

Two autonomous agents that find customers and capital for businesses that cannot afford a sales team. $0.50/day to run. Fortune 500 GTM at 0.05% of the cost.
$199K
Collected
$8K
Peak Day
$2.9M
ARR Velocity
$0.50
Cost / Day
0
Sales Reps
Book a demo — we run it on your prospects Try PKM Analyzer free
No deck. No slides. We show you the agent running in your terminal, on your ICP, in real time.
Live output

This is what AROS does every morning

No human. No CRM. Just the agent, running autonomously at 6am, for $0.50.

aros.py — daily autonomous run — March 24, 2026
$ python aros.py [06:01] Scanning SerpAPI for ICP signals... Found 23 home care agencies hiring caregivers Found 8 cleaning franchises expanding locations Found 4 HVAC companies with review spikes [06:02] Scoring 35 prospects by ICP fit... Tier A: 7 (score 85-100) | Tier B: 12 (60-84) | Tier C: 16 (skipped) [06:02] Running PKM defense profiling... Brian Torres — BrightClean Home Care, AZ Mode: OVERLOAD_AVOIDANCE (awareness: 3/10) Bypass: ULTRA_SHORT — 60 word hard cap Banned: "hope this finds you", "quick call", "at your convenience" Sarah Kim — CEO, CareFirst Solutions, NY Mode: IDENTITY_THREAT (awareness: 7/10) Bypass: AMPLIFICATION — "your judgment runs this" Banned: "let us fix", "you need", "struggling with" [06:03] Generating messages via Claude Haiku... Brian → 58 words, 0 banned phrases, bypass applied ✓ Sarah → 74 words, 0 banned phrases, bypass applied ✓ [06:03] Sending 19 emails via Gmail SMTP... ✓ 7 Tier A sent · ✓ 12 Tier B sent · All tracked in Airtable [06:04] Checking yesterday's replies... ★ HOT REPLY: Mike Rodriguez (CleanPro TX) "Tuesday 2pm works. Send me the link." → Meeting booked. Outcome logged. Pattern stored. Today: $0.47 cost · 19 emails · 1 meeting · $4,788 pipeline added
aria.py run — investor pipeline
$ python aria.py run [1/7] Finding investors... 148 found → filtered to 34 Tier 1 (thesis match 85%+) [2/7] Verifying emails... 30 valid · 4 bounces caught before sending [3/7] PKM defense profiling... Sarah Chen (Partner, Bessemer) → MOTIVE_INFERENCE (9/10) Bypass: PURE_DATA — open with number, never "I" [4/7] Writing (PKM applied)... "$199K collected. $8K peak day. $0.50/day. Zero sales reps. Code is public. You backed [co] — autonomous GTM at SMB layer. This is that, built and verified. 20 min. No deck." [5/7] Campaign created on Instantly... ✓ 19 emails queued · timezone optimized · reply tracking on [6/7] Synced to Airtable CRM ✓ Funds: General Catalyst, Bessemer, Antler, Blume, Kae Capital + 14 Raising: $250K pre-seed · Meetings: booked ✓
The problem

Two things every business needs.
Neither solved at scale.

$2M+

What Fortune 500 pays per year for GTM

40-person team. 100 tools. SDRs, AEs, CSMs, RevOps, demand gen. All of it to do two things: find revenue and find capital.

400M

Businesses that cannot afford any of it

They run on instinct, referrals, and luck. New revenue is their hardest problem. Capital to fuel growth is their second hardest. Both unsolved.

The solution

Two agents. One brain. One mission.

Same architecture. Same brain. One finds customers. One finds capital. Both learn from every outcome.

AROS
Autonomous Revenue Operating System
1
Discover prospects
Real-time signals — hiring, expanding, struggling
SerpAPI
2
Score by ICP fit
Trained on 4 months of real customer outcomes
scorer.py
3
Profile defense mode
10 modes detected from digital profile
PKM Engine
4
Write bypass message
Mode-specific strategy, banned words enforced
Claude Haiku
5
Send autonomously
Tracked, replies classified, outcomes logged
Gmail SMTP
6
Learn from outcomes
What worked → brain → next message smarter
Airtable
ARIA
Autonomous Relationship Intelligence Agent
1
Find investors
By thesis, stage, check size, portfolio signals
Apollo API
2
Verify emails
Catch bounces before sending — protect domain
Millionverifier
3
Research public activity
Portfolio moves, fund news, tweets
SerpAPI
4
Profile defense mode
VCs = MOTIVE_INFERENCE (awareness 9/10)
PKM Engine
5
Write & send campaign
PURE_DATA bypass, no pitch language
Claude + Instantly
6
Classify replies & alert
INTERESTED / OBJECTION / REFERRAL / NOT_NOW
Claude + HOT alerts
The perception layer — self-improving

PKM — gets smarter with every analysis

10 defense modes based on Friestad & Wright (1994). Every user who analyzes a profile and corrects the result trains the next classification. The more people use it, the more accurate it gets. A living system, not a static model.

Without PKM — deleted instantly
"Hi Brian — hope this finds you well. I'm really excited to share what we're building. Would love to set up a quick call to discuss how we can help grow your business."
4 banned phrases detected — triggers OVERLOAD_AVOIDANCE
With PKM — meeting booked
"Brian — 3 home care agencies in New York added 11 clients last month using Aonxi. You are hiring caregivers right now. 15 minutes Tuesday 10am?"
ULTRA_SHORT bypass applied — 28 words, 0 banned phrases, 1 calendar slot
Defense modeWhoBypass
MOTIVE_INFERENCEVCs, ex-GTM execs (Gong, HubSpot)PURE_DATA
TACTIC_RECOGNITIONRepeat founders, serial operatorsSIGNAL_HOOK
OVERLOAD_AVOIDANCESMB CEOs, home care ownersULTRA_SHORT
SOCIAL_PROOF_SKEPTICISMCTOs, engineers, technical leadsCREDIBILITY_FIRST
AUTHORITY_DEFERENCEDirectors, mid-level managersPEER_PROOF
LOSS_AVERSIONRisk-aware buyers, regulated industriesCOST_OF_INACTION
IDENTITY_THREATSelf-made founders, bootstrappedAMPLIFICATION
TIMING_SKEPTICISMPerpetually "not ready" buyersTRIGGER_EVENT
COMPLEXITY_FEARNon-technical ownersSIMPLICITY_PROOF
PRICE_ANCHORINGBudget-constrained buyersROI_INVERSION
Try PKM Analyzer free — your correction trains the next result
Every thumbs up/down and mode correction feeds back into the classification engine. View source on GitHub
The brain

Airtable — living intelligence
that compounds forever

Not a CRM. A pattern library. What worked for a home care agency in Arizona last Tuesday. What defense mode was cracked and how. Never forgets.

Agent sends message

Prospect, defense mode, bypass strategy, message, timestamp

Outcome recorded

Open, reply, meeting booked, deal closed, churned — every event

AIRTABLE

Shared between AROS and ARIA. Every customer makes the system better for every other customer.

Every
Contact
Every
Message
Every
Outcome
Every
Defense Cracked
Every
Pattern
Forever
Compounding

Next message is smarter

40 customers × 6 months × every defense mode = the agent today is better than yesterday

Proof

Every number real. Every commit dated.

$199K
Gross Collected
Oct 2025 — Mar 2026
$8,000
Peak Single Day
March 2026
$2.9M
ARR Velocity
That day annualized
$0.50
Cost Per Day
To run AROS
40
Customers
Home care, cleaning, franchise
$0
Raised
100% bootstrapped
38 repositories. 2 cores. 1 brain.

Every agent, every tool, every research paper
feeds into AROS or ARIA

Nothing exists in isolation. Every repo is a capability that makes one of the two cores smarter. New agents don't create new products. They merge in.

AROS
Revenue Intelligence Core — every repo below feeds into AROS
aros-agent
The core. Signal detection, ICP scoring, PKM profiling, autonomous email, daily briefings.
View repo →
CORE
aonxi-outreach-agent
Hyper-personalized outreach engine. 120+ persona models. The messaging layer AROS uses.
View repo →
PRODUCTION
pkm-analyzer
Defense profiling. 10 modes. Shared perception layer for both cores.
Try it live →
LIVE TOOL
simplenursing-agent
Firehose intelligence agent for SimpleNursing. Vertical-specific AROS deployment.
View repo →
CLIENT
simplenursing-gtm-engine
GTM intelligence engine for SimpleNursing. Full pipeline from discovery to close.
View repo →
CLIENT
myhq-gtm-engine
GTM engine for myHQ. Coworking vertical. Prospect scoring + outreach automation.
View repo →
CLIENT
attentive-gtm-agent
GTM engine for Attentive. SMS commerce vertical. Signal-based prospecting.
View repo →
CLIENT
bre-gtm-engine
GTM engine for BRE. Real estate vertical. Market intelligence + outreach.
View repo →
CLIENT
competitive-intelligence-agents
Competitive intel across markets. Feeds signals into AROS scoring decisions.
View repo →
INTELLIGENCE
b2b-funding-scanner
Real-time seed funding scanner. Finds newly funded companies for AROS to prospect.
View repo →
SIGNALS
news-briefing-agent
Bias-aware news briefing. Feeds market intelligence into AROS daily briefings.
View repo →
INTELLIGENCE
colibri-qa-agent
QA agent for Colibri. Conversation intelligence that trains AROS email quality.
View repo →
CLIENT
ARIA
Capital Intelligence Core — every repo below feeds into ARIA
ARIA
The core. Investor finding, thesis scoring, email verification, PKM bypass, reply classification.
View repo →
CORE
job-hunt-agent
Relationship intelligence for job searches. Same thesis-matching architecture as ARIA.
View repo →
RELATIONSHIPS
stock-analyst-agent
Market analysis agent. Financial intelligence that informs ARIA investor conversations.
View repo →
INTELLIGENCE
SHARED
Capabilities used by both AROS and ARIA
pkm-analyzer
Defense profiling — 10 modes. The perception layer both cores use before writing.
Try live →
PERCEPTION
memorymesh
Structured memory for agents. How both cores remember across sessions.
View repo →
MEMORY
echo
Agent communication layer. How AROS and ARIA share signals and outcomes.
View repo →
COMMS
glassbox-agents
Transparent agent execution. Every decision both cores make is inspectable.
View repo →
OBSERVABILITY
alignment-auditor
Alignment verification. Ensures both cores stay within safety and ethical bounds.
View repo →
SAFETY
realtime-reasoning-engine
Real-time reasoning for live decisions. Powers both cores' scoring and classification.
View repo →
REASONING
llm-evals-arena
LLM evaluation framework. How we choose which model powers which agent capability.
View repo →
EVALS
llm-calibration-study
Confidence calibration on B2B sales email quality. Ensures both cores know when they're wrong.
View repo →
RESEARCH
blueprint-trust-engine
Trust scoring for agent decisions. Determines how much autonomy each core gets.
View repo →
TRUST
RESEARCH
Papers replicated and deployed into production — the science behind the agents
asm-replication
Active Structured Memory. How agents remember what worked across months of campaigns. NeurIPS 2026.
View research →
NEURIPS 2026
prm-replication
Process Reward Models. Rewards intermediate reasoning steps, not just final outcomes.
View research →
PUBLISHED
rewardflow-replication
Multi-step reward propagation. Credits decisions that caused outcomes 47 steps later.
View research →
PUBLISHED
tdad-replication
Trajectory-Driven Agent Design. Learning from full action sequences, not single steps.
View research →
PUBLISHED
moe-efficiency-study
MoE routing strategy. Original metric. Directly applicable to multi-agent persona routing.
View paper →
PUBLISHED
information-architecture-thesis
Long-horizon state management. What to keep, compress, or discard over months of operation.
View thesis →
PUBLISHED
reward-model-blindness
RLHF reward models predict human ratings — but humans rate for the wrong reasons. Measures how wrong.
View research →
RESEARCH
reversion-study
Agent reversion patterns. Why agents regress after improvement — and how to prevent it.
View research →
RESEARCH
attention-from-scratch
Attention mechanism built from first principles. Foundation understanding that powers everything.
View repo →
FOUNDATIONS
META
The mission layer — vision, daily log, AGI architecture
agi-possible-agent
Daily AGI architecture experiments. Testing what's possible at the frontier.
View repo →
ACTIVE
frontier-agi-journey
Daily research log. Every paper read, every experiment run, every insight documented.
View log →
DAILY
space-wonder-agent
Exploration agent. Testing agent architectures in unbounded domains.
View repo →
EXPERIMENTAL
↓ ↓ ↓
All 38 repositories converge into
AROS
finds revenue
+
ARIA
finds capital
AIRTABLE
single source of truth — compounds forever
Coming next

LLM Content Layer

When someone asks any AI about your market, your business should be in the answer. Not SEO for 2015. Presence at the inference layer.

Claude
GPT
Gemini
Perplexity
Grok
Transparent pricing

3 months to discover your Revenue-Making Campaigns.
9 months to scale. You pay only from closed revenue.

We find the qualified meetings. We bring signal-driven prospects to your door. AROS discovers campaigns that convert. Then we fund the scaling ourselves.

Month 1-3
RMC Discovery Phase
Record and analyze 100% of sales calls. Generate 10-20 campaign concepts. Validate with 90-day performance data. You pay a flat monthly fee.
Month 4-12
Intelligent Scaling
Pay only from closed deals. We fund 100% of ad spend across all channels. Scale validated RMCs. Expand to new markets. Revenue share only.
Starter
1 Revenue-Making Campaign
Month 1-3: Discovery
$3K/mo
Month 4-12: Scale
8-12% revenue share
on closed deals only
1
Guaranteed RMC
$30K+
Capital Deployed
  • 1 validated RMC guaranteed
  • Google Ads + Meta Ads testing
  • AI sales intelligence from 100% of calls
  • PKM defense profiling on every prospect
  • Live campaign ledger tracking
  • $30K capital deployed after validation
  • Multi-channel expansion in scaling phase
Get started
Scale
5 RMCs for Domination
Month 1-3: Discovery
$10K/mo
Month 4-12: Scale
8-25% revenue share
on closed deals only
5
Guaranteed RMCs
$100K+
Capital Deployed
  • 5 validated RMCs guaranteed
  • All channels: Google, Meta, YouTube, Display, LinkedIn
  • AI sales intelligence from 100% of calls
  • PKM defense profiling on every prospect
  • Live campaign ledger tracking per RMC
  • $100K capital deployed after validation
  • Aggressive multi-channel expansion
  • Fastest iteration cycles (1-2 days)
  • Multi-location support
  • Dedicated success manager
Get started
How it works: AROS discovers signal-driven prospects and qualified meetings. PKM profiles every prospect's defense mode before any outreach. You only meet people who are already showing buying signals — hiring, expanding, struggling with a problem you solve. After validation, we fund the ad spend ourselves. You pay revenue share on closed deals only. If we don't close, you don't pay.

The guarantee

Validated Revenue-Making Campaigns within 90 days — or we keep working for free. We can guarantee this because AROS has already collected $199K at $0.50/day. The patterns are proven. Every new customer adds more patterns.

We find the meetings. We bring the prospects.
You close the deals. We scale the winners.

Book a call — see AROS find prospects for your business
origin@aonxi.com · aonxi.app · github.com/originaonxi