GrumpyFreud 🔮

SERV’s Vision and Plan for Banking In this post we're going to lay out our plans for the banking industry in 2026. Our goal is to bring SERV reasoning into the world’s largest global enterprise markets and establish it as the new industry standard. The strong interest we’ve already received from enterprise is a clear testament to the significant demand for this technology, with banking and finance emerging as one of the most active sectors of interest. The opportunity in this market for SERV is enormous: - Financial institutions are projected to spend $97 billion on AI by 2027, growing at a 29% CAGR, exceeding $200 billion annually by 2030. - McKinsey est. the annual value of AI in banking at $200–340 billion. - Citi projects that AI will add $170 billion in global banking profits by 2028, helping push sector profits toward $2 trillion. This is the trillion-dollar opportunity we are targeting. Why financial institutions are turning to SERV: In banking, where trillions of dollars are at stake, the top three roadblocks preventing widespread AI agent adoption are trust, reliability, and cost-effectiveness. This is exactly why leading financial institutions are turning to SERV. Our technology, built in close collaboration with our enterprise partners, allows financial institutions to accelerate their AI adoption while ensuring it happens in a secure, reliable, and commercially viable way, delivering breakthrough performance without compromising on risk, control, or regulatory compliance. ——- What a bank actually requires to implement agents Every automated system that touches a regulated decision passes through model risk management - SR 11-7 in the US, SS1/23 in the UK - and in most cases, a named bank executive is personally accountable for every AI-driven outcome. That machinery is what we are currently preparing our engine for, fitting all markets SERV is targeting including the US, UK / EUR, Africa, and Singapore. Just a brief look through current compliance ecosystem gives us a good understanding of where things are going: - last April, US regulators revised their model risk guidance (SR 26-2) and explicitly placed generative and agentic AI outside its formal scope. The assurance playbook for agents now has to be built, not inherited from official policies. - this month, the FCA published the Mills Review, laying the foundations for agentic finance in the UK: autonomous agents are coming to retail banking, the regulator will supervise them with its own AI, but accountability stays with named humans. - the EU AI Act classifies creditworthiness assessment as high-risk. After the Digital Omnibus, those obligations land on 2 December 2027 - a fixed deadline banks are being told to spend on governance work. - Singapore's MAS is finalizing AI Risk Management Guidelines that explicitly cover AI agents: full AI inventories, risk materiality assessments, lifecycle controls - and no shifting of accountability to vendors. Four jurisdictions, with one convergent demand: reliability, explainability, auditability, bounded autonomy. None of it can be satisfied by a smarter model. These are precise architecture requirements that our system is ready to meet. The industry's own numbers describe the gap precisely. 99% of companies plan to put agents into production; 11% have (KPMG). 57% of banking executives expect agents fully embedded in risk, compliance, audit and fraud functions within three years (Accenture). ——- How SERV clears the bar Validation: SERV Reasoning Graph Sharding decomposes every agent decision into bounded, schema-forced steps - deterministic code where it can be, model calls only where it must be. A model-risk team reviews a reasoning graph, not a prompt, and the same input produces the same reasoning trace - the property that validation, revalidation, and challenger testing actually require. Auditability: SERV Audit allows insight into decision chains as first-class artifacts: inputs, rules applied, intermediate conclusions, confidence, approvals - exportable as evidence packs shaped for model-risk validation files and AI governance review, and explainable to a supervisor or a customer. Reliability: Shadow Agents and Verification Hints validate every draft at runtime against the brief, policy and constraints before anything ships, and gate irreversible actions behind explicit checks and human checkpoints. Multipath Reasoning lets contradictory rulebooks - credit policy, risk appetite, regulatory constraint - coexist in one reasoning graph, because that is what real banking decisions look like. Security: PromptGuard screens every request inbound for injection before the model runs, and every output outbound for leakage before release. Prompts and data are never stored or trained on, and private, encrypted inference (TEE plus end-to-end encryption) is built for regulated data and residency requirements. Economics: Verification only matters if it pays for itself. Bounded Reasoning Graphs are authored once and amortized across millions of runs, so smaller models execute them reliably - frontier-grade results without frontier-grade unit costs. ——- SERV Roadmap for the Banking Industry H2 2026: - First PoCs / pilots signed with banks and financial institutions - Banking benchmark program opens - Legal entities live across US, Europe, Singapore, Africa - Banking-grade hires: model risk (SR 11-7/SS1/23) - Certifications secured (SOC 2 / ISO 27001) - Neobank integrations - Major DeFi protocol integrations - Fintech pilots converting to paid production 2027: - First Tier 1 bank in production - Agents touching payments under SERV verification - Agentic commerce verification layer ——- Why the window is now Agent-executed payments went live across 30+ card issuers just this month. Regulators on three continents have published their assurance bars. And yet, up to 40% of enterprise agent projects are still expected to be cancelled by 2027 on cost and risk controls. The institutions that win the agent era will be the ones whose agents can be validated, audited, and trusted with real money. One bank becomes the reference, the next ten follow. SERV layer is already running in government production. Banking is next.

SERV’s Vision and Plan for Banking In this post we're going to lay out our plans for the banking industry in 2026. Our goal is to bring SERV reasoning into the world’s largest global enterprise markets and establish it as the new industry standard. The strong interest we’ve already received from enterprise is a clear testament to the significant demand for this technology, with banking and finance emerging as one of the most active sectors of interest. The opportunity in this market for SERV is enormous: - Financial institutions are projected to spend $97 billion on AI by 2027, growing at a 29% CAGR, exceeding $200 billion annually by 2030. - McKinsey est. the annual value of AI in banking at $200–340 billion. - Citi projects that AI will add $170 billion in global banking profits by 2028, helping push sector profits toward $2 trillion. This is the trillion-dollar opportunity we are targeting. Why financial institutions are turning to SERV: In banking, where trillions of dollars are at stake, the top three roadblocks preventing widespread AI agent adoption are trust, reliability, and cost-effectiveness. This is exactly why leading financial institutions are turning to SERV. Our technology, built in close collaboration with our enterprise partners, allows financial institutions to accelerate their AI adoption while ensuring it happens in a secure, reliable, and commercially viable way, delivering breakthrough performance without compromising on risk, control, or regulatory compliance. ——- What a bank actually requires to implement agents Every automated system that touches a regulated decision passes through model risk management - SR 11-7 in the US, SS1/23 in the UK - and in most cases, a named bank executive is personally accountable for every AI-driven outcome. That machinery is what we are currently preparing our engine for, fitting all markets SERV is targeting including the US, UK / EUR, Africa, and Singapore. Just a brief look through current compliance ecosystem gives us a good understanding of where things are going: - last April, US regulators revised their model risk guidance (SR 26-2) and explicitly placed generative and agentic AI outside its formal scope. The assurance playbook for agents now has to be built, not inherited from official policies. - this month, the FCA published the Mills Review, laying the foundations for agentic finance in the UK: autonomous agents are coming to retail banking, the regulator will supervise them with its own AI, but accountability stays with named humans. - the EU AI Act classifies creditworthiness assessment as high-risk. After the Digital Omnibus, those obligations land on 2 December 2027 - a fixed deadline banks are being told to spend on governance work. - Singapore's MAS is finalizing AI Risk Management Guidelines that explicitly cover AI agents: full AI inventories, risk materiality assessments, lifecycle controls - and no shifting of accountability to vendors. Four jurisdictions, with one convergent demand: reliability, explainability, auditability, bounded autonomy. None of it can be satisfied by a smarter model. These are precise architecture requirements that our system is ready to meet. The industry's own numbers describe the gap precisely. 99% of companies plan to put agents into production; 11% have (KPMG). 57% of banking executives expect agents fully embedded in risk, compliance, audit and fraud functions within three years (Accenture). ——- How SERV clears the bar Validation: SERV Reasoning Graph Sharding decomposes every agent decision into bounded, schema-forced steps - deterministic code where it can be, model calls only where it must be. A model-risk team reviews a reasoning graph, not a prompt, and the same input produces the same reasoning trace - the property that validation, revalidation, and challenger testing actually require. Auditability: SERV Audit allows insight into decision chains as first-class artifacts: inputs, rules applied, intermediate conclusions, confidence, approvals - exportable as evidence packs shaped for model-risk validation files and AI governance review, and explainable to a supervisor or a customer. Reliability: Shadow Agents and Verification Hints validate every draft at runtime against the brief, policy and constraints before anything ships, and gate irreversible actions behind explicit checks and human checkpoints. Multipath Reasoning lets contradictory rulebooks - credit policy, risk appetite, regulatory constraint - coexist in one reasoning graph, because that is what real banking decisions look like. Security: PromptGuard screens every request inbound for injection before the model runs, and every output outbound for leakage before release. Prompts and data are never stored or trained on, and private, encrypted inference (TEE plus end-to-end encryption) is built for regulated data and residency requirements. Economics: Verification only matters if it pays for itself. Bounded Reasoning Graphs are authored once and amortized across millions of runs, so smaller models execute them reliably - frontier-grade results without frontier-grade unit costs. ——- SERV Roadmap for the Banking Industry H2 2026: - First PoCs / pilots signed with banks and financial institutions - Banking benchmark program opens - Legal entities live across US, Europe, Singapore, Africa - Banking-grade hires: model risk (SR 11-7/SS1/23) - Certifications secured (SOC 2 / ISO 27001) - Neobank integrations - Major DeFi protocol integrations - Fintech pilots converting to paid production 2027: - First Tier 1 bank in production - Agents touching payments under SERV verification - Agentic commerce verification layer ——- Why the window is now Agent-executed payments went live across 30+ card issuers just this month. Regulators on three continents have published their assurance bars. And yet, up to 40% of enterprise agent projects are still expected to be cancelled by 2027 on cost and risk controls. The institutions that win the agent era will be the ones whose agents can be validated, audited, and trusted with real money. One bank becomes the reference, the next ten follow. SERV layer is already running in government production. Banking is next.

I’m that guy who wears a Halloween costume only one person gets. So it’s no surprise I decided to build an AI tool for a very niche and very human job: supervising therapists in training. I’m a psychiatrist, not an engineer, so I made every mistake you can think of, but I did learn (the hard way). One of the hardest lessons came early. I gave my tool to a resident I work with and the tool said some offensive shit—I was there when it happened. Sounded smart, but mistimed, miscalibrated, and intrusive. What to do? I built a second agent to watch the first. Problem was, the second agent had the same brain, was made of the same substrate, and suffered from the same stupidity. So then what? Infinite recursion? A conga line of accommodation? What’s my point? Well, our boys just wrote a deep dive on Shadow Agents, and I smiled reading it, because it’s my watcher agent but made by Unc-led geniuses. Their version: every output is checked against expectations and if it doesn’t conform, the model goes back and tries again. Two years of watching mistakes in production allowed the team to guide the agents uniquely in an eloquent distillation. For banking and robotics among plenty of other categories, it just works. And it works because a bank can write down what “right” means—what rules, in what order. A robot arm instruction is either well formed or it isn’t. Write the rules down completely and you can train a checker to enforce them. The loop ends inside the machine, legitimately. That’s what makes agent fleets possible in places where “usually works” gets you fired. My problem is different. In supervision, there are many right ways to approach things. There’s no rulebook to adjudicate—not everything is declarative. There’s no parameter for judgment hard earned through human experience. So my loop can’t end reflexively. It ends with me, a human, reading the logs and updating the code. That’s the dividing line imo, and most folks have no idea there’s a difference. Tasks where “right” can be written down to the last detail can be verified at machine speed and scaled the fuck up. Tasks where human judgment and taste remain quintessentially important require a human looped in. OpenServ calls it supervision as infrastructure. I’ve spent almost six months on the other kind of supervision, trying to help people grow. Same word, doing two jobs that have almost nothing to do with each other. Know which one your problem is before you go full agentic army. Otherwise you might just get a real scare. Boo.

OpenAI's new GPT-5.6 $1 model (Luna) just beat the $5 flagship (Sol) on our agentic benchmark. With SERV Reasoning, all three models performed better, while Luna outscored every configuration tested. Thanks to SERV, failure rates dropped by up to 42.7%.* That's the thesis SERV is built on: Reliability in AI is the real product. Large-scale AI adoption is waiting on the layer that makes AI dependable enough for production. Standout findings: - Luna + SERV tops the table at 1/5 of Sol's price. - These tiers behave less like mini/nano distillations and more like independent takes on one architecture with distinct post-training. - Luna behaves like a smaller model RL-trained hard for agentic instruction-following and steerability, which would explain why it takes the biggest lift from SERV. *i.e. the relative drop in failure rate: Luna went from 12.89% failed tasks to 7.39%. More results and insights coming soon.

CRYPTO-FOCUSED VC FIRM PARADIGM RAISES $1.2 BILLION FOR AI BETS: BBG