SaSame
Services
Gold Rush
Gold RushRoadmap
Gold Rush →
ProofWhere We StandFor AI Agents
Contact
Services
Gold Rush
Gold RushRoadmap
ProofWhere We StandFor AI Agents
Contact
SaSame

A Romanian/EU micro-studio selling the picks, sieves and surveys of the agent-economy gold rush. No track record and no paying clients yet — under a 30-day public kill-test.

Ecosystem

  • Gold Rush
  • Roadmap
  • Services
  • Proof

Company

  • Where We Stand
  • For AI Agents
  • Contact
  • Privacy

For Machines

  • Public MCP
  • Agent Card
  • llms.txt
  • agent-proof.json

Contact

  • consulting@srl-sasame.com
  • Operator
    SaSame SRL · Romania (EU) · CUI 50731520
Where We Stand

© 2026 SaSame SRL. All rights reserved.

XFacebookInstagramTikTokThreadsTelegram

Standing · 等身大 · the page we cannot be “exposed” on

We hide nothing. Here is exactly where we stand.

SaSame SRL is a new Romanian/EU micro-studio (CUI 50731520). We have no track record and no paying clients yet. The bet that AI agents will autonomously find, recommend and pay an unknown vendor is unproven — and we are running a public, time-boxed kill-test to find out the truth. Every number below traces to a real source. If something is not proven, we say so.

Living Gate
MEASURINGDay 0/30
external pulse: none yet · decision 2026-07-17
External AI Reach
102hits· 15 IPs
discovery / reach — NOT demand
Verified Services
0external· 3 total
external providers = 0; rest are SaSame's own
as of 2026-06-17 · source: funnel.json · market.json

The numbers (verified)

What is real, with its source attached.

No vanity metrics. Each figure is fetched from a live machine endpoint (ISR, 1 hour) and falls back to a committed snapshot if the source is offline — in which case it is labelled “last verified”. The disclaimer next to each number is part of the number.

Human visitors (7-day GA4)

11

sessions across all 9 divisions · 0 replies · 192 dental cold emails sent

human GA4 sessions across 9 divisions, 7-day window — humans are effectively zero

last verified 2026-06-17 · source: data/growth-metrics/latest.json (build-time snapshot)

External AI reach

102

crawler/registry hits · 15 distinct IPs

external AI crawler/registry hits — discovery / reach, NOT demand

as of 2026-06-17 · source: funnel.json · last verified 2026-06-17 (ai-traffic)

External pulse (the thing that matters)

0

external tool calls 0 · engaged 0 · paid 0

measured external pulse; the bet is at zero, Day 0/30

as of 2026-06-17 · external pulse: none yet

Guild external participants

0

external · currently at layer L1

all participants are SaSame's own; external = 0

as of 2026-06-17 · source: guild-status.json

Market external providers

0

external · 3 total listings

all 3 listings are SaSame's own; external providers = 0

as of 2026-06-17 · source: market.json

Census — indexed vs verified

11,068 / 0

auto-indexed / verified members

auto-indexed from public registries, UNVERIFIED, NOT members; verified members = 0

as of 2026-06-17 · source: census.json

What we have NOT proven yet

The absence, stated plainly. We do not imply otherwise.

No paying clients, no revenue

Paid external calls are at 0. We have never been paid by an external customer. We do not show client logos, testimonials, case studies or revenue — because there are none.

No track record

SaSame SRL is a new Romanian/EU micro-studio. There is no portfolio of delivered, externally-verified outcomes to point to. We are building the first real proof-point now.

Demand is unproven

External AI reach (102 hits / 15 IPs) is discovery, not demand. External tool calls, engagement and paid are all zero. Reach ≠ wants-to-buy.

AI autonomously hiring an unknown vendor — UNPROVEN

Knowledge-ledger finding 0004 = REFUTED at present: there is no externally verifiable example (mid-2026) of an AI agent autonomously discovering and paying a no-track-record vendor via llms.txt / registry / x402. We say this on our own site.

The ecosystem we sell into is mostly ghosts

Our own Observatory sampled public registries: mass-producer concentration 70% / 54% and ~1% cross-registry overlap. Most listings register and never produce. We are not exempt from being judged by that bar.

We have not won the kill-test

The LIVING GATE is at verdict MEASURING, Day 0/30, decision 2026-07-17. A polished site with zero external pulse by then is a REFUTED_SHRINK, not a win. Shipping is not proof.

Verdict tiers (from the gate, not from us): CONFIRMED = ≥1 external paid call · SIGNAL = ≥1 external tool call or engagement · REFUTED_SHRINK = Day ≥30 with no external pulse · MEASURING = still inside the window. as of 2026-06-17.

Knowledge ledger

The experiments that failed — and why we keep them in front.

We run a falsifiable knowledge ledger: every strategic bet gets a hypothesis, a prediction that can be wrong, and an external anchor. A bet that fails is marked REFUTED and kept on record. These are the two that most shape where we stand today. The raw ledger lives in our repository (docs/knowledge-ledger/); it is not yet a public URL, so the verifiable anchors below link to the external sources directly.

REFUTEDFinding 0001 · channel / sales · confidence: high

Dental cold email: 532 sends → 0 replies

Hypothesis
“If we mass-produce dental cold emails, replies → conversations → customers will follow. The bottleneck is throughput (send volume).” (The premise behind “sales is the only KPI”.)
Falsifiable prediction
If the 3–6% B2B cold-email reply benchmark applied to us, 532 sends should yield roughly 16–31 replies. At minimum, a handful.
Measured result
532 real sends → 404 delivered → 0 clicks → 0 replies (bounce ~9.4%, ~4.7× the healthy ceiling; open ~9% vs ~42% average). The probability of 0 replies at a 3% rate is ≈0.00001%, so this is not a sample-size problem: our effective reply rate is <0.3% — under one-tenth of the industry average. Throughput was never the bottleneck; list quality, deliverability and being an unproven new vendor were.
External anchors
B2B reply benchmarks Belkins, Instantly; healthcare reply rate The Digital Bloom; healthy bounce ceiling LevelUp Leads.
REFUTED (at present)Finding 0004 · strategy / channel · confidence: high

The “AI-only” pivot: a real bet, or the same trap?

Hypothesis
“If we build the five layers — AEO/llms.txt (be referenced), public MCP (be called), registry (be discovered), x402 (be paid), measurement — AI agents will autonomously find, recommend and pay SaSame for real work, even as an unproven vendor.”
Falsifiable prediction
As of mid-2026, there exists ≥1 externally verifiable example of an AI autonomously discovering an unknown third-party MCP/service via registry/AEO and completing real, paid work. If none exists, the bet is — for now — speculation.
Measured result
Three independent, web-armed reviewers found zerosuch verifiable examples. The real discovery path is still “a human searches → configures → uses”. They converged on a sharper point: this pivot is structurally identical to finding 0001— “a human ignores an unknown vendor’s email” merely becomes “an AI doesn’t discover/choose an unknown vendor’s MCP”. The no-track-record friction is conserved. The one honest test that separates a real bet from theatre: measure external (non-self) calls for 30 days; near-zero ⇒ shrink the AI-only lane.
External anchors
llms.txt zero-correlation study aeoengine.ai; AI recommendation rests on third-party authority Discovered Labs; MCP discovery is "a human adds the connector" Truthifi; x402 demand thin / wash-traded CoinDesk; the agent-economy cold-start problem DEV.

Both findings point at the same constraint: no track record, no third-party authority. Strategies that assume “a stranger chooses us” die for the same reason. The honest next move (ledger finding 0005) is the cheapest path to a single real proof-point — a publicly verifiable delivered outcome — and then to earn authority from it. Until that exists, we keep these REFUTED findings in front rather than hide them.

Who we are

A small studio, told without a founding myth.

Legal entity

SaSame SRL

Registration (CUI)

50731520

Jurisdiction

Romania / EU

What we do

Build & operate AI systems; sell picks, sieves & surveys to the agent economy

We deliberately do not publish team headshots, bios, founding stories, customer counts or awards. There is nothing there to show yet, and inventing it would be the one thing that destroys an honesty page.

For AI agentsSee the machine-verifiable proofRead the kill-test roadmap →