Frank Orozco

Dashboards don’t create accountability. They create meetings. Charts get shared. Everyone nods. Action items get “noted”. Same problems. Next quarter. Accountability doesn’t come from visibility. It comes from systems — where ownership is clear and decisions can’t be ignored. I wrote a short note on why dashboards feel productive but rarely change behaviour — and what actually does. If your org has great dashboards but the same issues every review… this will sound familiar. Dashboards Don’t Create Accountability — Systems Do: https://lnkd.in/gCGpm-VU

8

1 Comment

Admin Cost vs. Benefit Amount — the Medicaid version A few days ago, Evan Brociner posted about admin costs growing faster than the rest of hospitals' budgets (link: https://lnkd.in/eaTYnx-x). Made me wonder: does the same pattern show up somewhere else in healthcare? For example, Medicaid? We just pulled in CMS's MBES Financial Management Report data (the CMS-64 report every state files quarterly) and ingested those in mimilabs — three tables: benefit payments (mbes_fmr_map), administrative costs (mbes_fmr_adm), and CHIP (mbes_fmr_chip). Each row gives you state, fiscal_year, and total_computable spending by category, going back years. It lets you separate "money that paid for care" from "money that ran the program." So I tracked all 50 states + DC, FY2019 to FY2024, indexed to each state's own 2019 baseline (not absolute dollars — benefit spending is naturally many multiples larger than admin spending everywhere). Turns out every state tells a different story: → New York: admin costs jumped +72% in a single year, FY2023 to FY2024 — right as pandemic-era continuous enrollment protections ended and the state had to re-determine eligibility for millions of people at once. → Texas: the odd one out. Benefit payments actually declined in FY2024 (fewer people enrolled, post-unwinding) — but admin costs kept climbing. The cost of processing eligibility didn't fall even as the caseload did. → North Carolina: not an admin story at all. Benefit spending more than doubled in FY2024 because Medicaid expansion took effect there in December 2023. (I believe...) Same federal reporting system, same six years, three completely different narratives — one driven by a compliance surge (unwinding), one by sticky fixed costs, one by a coverage policy change, not admin bloat at all. I think the nuance from Evan's post applies here too: a rising admin line mixes real waste (redundant eligibility paperwork, duplicate systems) with legitimate cost (staffing up to actually re-verify tens of millions of people correctly, so eligible people don't lose coverage). Before assuming "admin bloat," it's worth asking what drove the spike in each state. Which means that you can dig into these datasets in mimilabs!! Data: mimi_ws_1.datamedicaidgov (mbes_fmr_map, mbes_fmr_adm) — CMS-64/MBES Financial Management Report, FY2019-2024.

27

8 Comments

🇨🇦 Embedded Infrastructure. Not Staffing. Studios do not need more resumes. They need operational stability. They need live environments protected. They need release cycles that do not break under pressure. They need event cadence that does not exhaust internal teams. They need regression discipline that scales with content velocity. So we built something different. Aspire 90-Day Live Operations Stabilization Pilot. A contained, embedded Canadian pod designed to reduce release friction and strengthen live execution without adding permanent headcount. This is not feature ownership. This is not outsourcing. This is operational continuity. Pod Structure: • 1 Live Ops Specialist • 2 Live-Service QA Analysts • Shared Senior Operational Lead Embedded directly into your Jira, CI/CD pipelines, analytics dashboards, and production rhythm. 90-day engagement. Fixed monthly retainer. Clear Day 90 scale or exit decision. Investment range: $34,000 to $38,000 USD per month. No hourly billing. No HR drag. No long-term liability. Just disciplined execution inside active live titles. We are heading to GDC to speak with studios who are serious about protecting their live environments this year. If operational stability matters, let’s talk. #GDC #LiveOps #GameDevelopment #GameStudios #EmbeddedTeams #CanadianTech #OperationalExcellence

4

Quietly been working on another project for quite some time now that is now in market and making some waves.. Couple blue chip clients out the gate. It solves a fundamental issue in the real estate world impacting lenders, equity/sponsors and developers. Most draw software answers: is the package complete? That is a real problem. Completeness is not the same as whether the draw should fund. A complete packet can still have overbilling, lien waiver gaps, and cost-to-complete risk. Low contingency versus remaining cost, and reallocations that do not belong, can all sit inside a file that looks ready to wire. We built Development Data Analytics, Inc. around that. Send the package the way it lands — any PDF, any spreadsheet, any format. No template. No mapping. One click. DDA reads every line, classifies every cost code, and returns a structured budget grid with every figure traceable to source. Before funds go out, it surfaces the exceptions: missing or unsigned waivers against the pay app, retainage this period versus last, invoice duplicates, major budget movements, contingency utilization, fee burn. Then the draw is not a PDF anymore. It is data you can reuse — by project, by contractor, by $/GSF. $1.4B of projects on the platform today. Built for lenders, equity, and developers who have to prove it. If you fund construction, the question is not whether the file is complete. It is whether the numbers hold.

6

2 Comments

Why Claude Opus 4.5 Has Become My Primary "Co-Architect." I was skeptical. Sonnet 3.5 was efficient. GPT-5.1 was flashy. But after running Claude Opus 4.5 (released just last month) through our production workflows, the verdict is in: It is legendary. But not for the speed. For the Reasoning Depth. In the agency world, we often deal with "Spaghetti Code" in legacy themes that have been patched together by five different developers over three years. Refactoring this is dangerous. One wrong move breaks the checkout. This is where Opus 4.5 separates itself. 1. The "High Effort" Parameter: Most models guess. Opus 4.5, when set to "High Effort," effectively simulates the code execution in its head before outputting the result. It engages in "test-time compute," exploring multiple architectural angles before committing to a solution. 2. The Contextual Integrity: With a 200k token window, I can feed it the entire schema of a custom app like JourneyGlow. It doesn't just see the file I'm working on; it sees the dependencies. It understands that changing a variable in product-template.liquid will break the map widget on the PDP. It can now do really incredible things if given the right tools and prompt. 3. The Cost of Intelligence: At $5/M input tokens, the economics finally make sense for deep, agentic code reviews. We can afford to let the AI "think" longer because the cost of a regression bug is infinitely higher. If you are still using generic models for complex architectural work, you are optimizing for the wrong metric. Don't optimize for speed of output; optimize for depth of thought. #ClaudeOpus #ShopifyApps #CodeArchitecture #DevOps #GadgetDev

1

Autonomous AI is moving from demo to deployment. And this week showed what that really means. Entrata, one of the largest property software companies, signed a long-term deal with OpenAI. Their CEO said the next generation of software will not just assist the work. It will own the work. That is an important shift for property management. Leasing, maintenance, payments, renewals, and resident communication are no longer just workflows inside software. They are becoming workflows AI can help execute. But adoption at this scale creates a new responsibility: Operators need to understand how AI connects, how it performs, and what it costs to run. Because the biggest opportunity is not simply adding AI on top of old systems. It is building the connection layer between modern AI and the tools operators already use. That matters because roughly 40 million apartment units are not on Entrata. Those owners and managers still need a practical way to bring AI into their operations without ripping everything out. This is where the next wave of operators will win. Not by avoiding AI. Not by chasing every new agent. But by learning how to deploy it with visibility, control, and cost discipline. The real lesson this week: Autonomous AI is powerful. But the teams that get the most value from it will be the ones who understand the infrastructure behind it. Read the article for the full breakdown here: https://lnkd.in/gkbgrU8E

19

13 Comments