I . The bar
Each lab provisions a live GCP project for the candidate. Vertex AI calls cost real money. Mistakes cost real budget.
The clock starts when you click "Start". It does not pause. If your sandbox bricks, support credits the time. There is no negotiation on the bell.
You cannot edit your diagnosis after you submit it. You cannot un-pick a trap. You cannot take back a bad executive note.
II . The launch line
One scenario goes live with the launch. Five join across Q3 and Q4 of 2026.
Eight production AI agents are burning $48,000 a day. The CFO wants 30 percent off in 45 minutes. Pick the right 5 actions. Defend them in writing.
A 600 node GKE fleet just absorbed a noisy neighbor and a misconfigured HPA. Costs are out of band. Stabilize the cluster, attribute the spend, and write the executive note.
Gas costs on a layer 2 rollup are eating margin. Validators are complaining. Identify the contract calls driving the burn, propose a settlement plan, and present to a treasury committee.
You inherited a developer platform that 600 engineers depend on. Costs grew 4x in 18 months while utilization plateaued. Build a FinOps program for an internal platform without breaking developer flow.
A misconfigured agent is logging full prompts to a public BigQuery table and spending $9,000 an hour. You have one hour. Stop the bleed. Preserve the audit trail. Write the public statement.
A Cloud Run plus Cloud Functions plus Workflows pipeline is overrunning budget by 60 percent. Find the cold starts, the over-provisioned concurrency, and the egress traps.
III . The tiers
The deterministic decision and execution scores set the floor. The LLM-graded diagnosis and executive scores can lift you. They cannot save a wrong action selection.
IV . Comparison
V . Scoring and grading
Three short answers, 60 seconds each. LLM scored against expected concepts and causal reasoning.
Up to five action selections from a mix of GOOD and TRAP cards. Deterministic. Traps drag you negative. Picking zero traps earns a bonus.
Sum of the dollar value of GOOD actions you selected. Mapped linearly against the required reduction.
Five short answers including a strategy choice. LLM scored on clarity, defensibility, and CFO-readability.
VI . The evolving badge
Each Proven cert you earn becomes a satellite medallion around the seal. The active cert sits at the center. Prior certs orbit at the cardinal points. The footer ribbon names your tier and the date.
Cross four Elite tier credentials in a calendar year and your seal lifts a gold leaf foil overlay across the whole badge. That is the Top 1 percent.
The badge is anchored to RING:1000:2026 and signed to Sigstore on issuance. Each medallion encodes the cert code, tier, and issuance date in the SVG's machine-readable footer.
VII . Who qualifies
You write the gcloud commands, you tune the LangGraph orchestrators, you watch the dashboards. Proven validates the muscle you already use.
You decide the routing layer, the caching layer, the observability stack. Proven asks you to make those decisions live, against a real ledger.
You answer to a CFO. Proven gives you a page of writing your CFO will read verbatim, and a credential they will recognize.
VIII . Access
Proven Labs
Unlimited live labs. Every cert. Every retake. AI graded. Evolving badge across the line.
On Free or Professional today. Upgrade to Executive to unlock Proven Labs.
Proven Exams
The formal Proven Exam track is the proctored, time-locked, single-attempt path that produces a verified credential. It has its own pricing model. Coming Q3 2026.
Explore certificationsIX . Recognition and standards
Every Proven badge is anchored to the IFO4 ring registry. The ring number plus the cert code form a verifiable lookup.
Our exam standards are certified to the Unified FinOps Methodology Standard. Each lab is mapped to UFMS competency lines.
Each issued credential is signed to Sigstore on mint. Anyone can verify the signature against the public key without contacting IFO4.
X . FAQ
No. IFO4 Proven is multi-cloud and multi-domain. The Agentic AI lab spans Vertex AI plus the Anthropic API. Future labs cover GKE, layer 2 chains, internal developer platforms, serverless, and incident response.
Yes. Each attempt requires a fresh exam ticket. Your previous attempts and badge layers remain in your public profile.
Your attempt state is checkpointed every action. If the sandbox is unrecoverable inside the timer, support pauses the timer and re-issues the attempt for the remaining minutes.
A deterministic rubric grades your action selections and the dollar value of your remediation. A Vertex AI Gemini 2.5 Pro examiner grades your written diagnosis and executive justification. The deterministic score is the floor.
A written exam tests recall. Proven tests practice. You sit in front of a real cloud sandbox in incident, you take action, and your decisions are scored against the bill the company would actually pay.
Badges are anchored to the IFO4 Standards Lab and to RING:1000:2026. They do not expire. New labs add satellite medallions to your existing badge instead of replacing it.
Yes. Each Proven credential mints into the IFO4 soulbound credential collection on Polygon. You can also share a public link without a wallet.
Proven Labs are free for IFO4 Executive Members. Free and Professional members upgrade to Executive to run unlimited labs. Proven Exams (the formal certification track) are a separate product with their own pricing.
Yes. Executive Membership covers Proven Lab access. Reach out to certifications at ifo4 dot org for enterprise group memberships.
One scenario. Live now.
45 minutes. Eight production agents. $48,000 a day. Pick the right 5 actions and defend them in writing.
Open the Agentic AI lab