All Posts

A recomputation of the Landis et al. 2021 sourdough census. Where the baker lives did not sort the starter.

After the package is held still, what happened to grocery prices in Denver, and where does the public record stop? Denver-Aurora-Lakewood food at home in August 2026 is an index of 288.763, up 1.7 percent over the year. The matching U.S. city average is up 2.2 percent. An index is not a shelf price, and it is not a King Soopers receipt. The dollar figure BLS does publish is a national average price, already per dozen, per gallon, or per pound. One dozen grade A large eggs, on

The Vision Zero page said 94 deaths in 2025. The open crash file records 41, and 0 pedestrian deaths.

16,281 federal contracts end in FY27. I built the missing ground truth from public award data and a calibrated model that predicts which ones change vendors, and it beat the heuristic in my own product.

I cleaned and analyzed 3.0 million Denver 311 contacts from 2019-2025. The biggest slowdowns in the data are record-keeping gaps, not service, and the fixes are operational.

Most AI overspend is not from using AI. It is from running a frontier model for work an economy model would do identically. The Modelnomics framework for matching model to task.

Most security advice for small companies sounds like it was written for a Fortune 500. This is what credential hygiene actually looks like for a 5-50 person team, shipped for a real client, sanitized and generalized.

An anonymized field case study Client details are withheld by request. This is a real engagement with a fast-scaling, multi-region direct-to-consumer brand. Names, products, internal documents, and proprietary figures have been removed or generalized. The numbers below are real and were measured against a 60-day daily work log. The short version A consumer brand brought me in on a contracted, single-operator pilot to do four things: rebuild the web and landing-page layer, own

Checklists and policy templates won’t save you under Colorado SB 24-205. The affirmative defense requires accumulated evidence over time — bias audits, consumer disclosures, incident response logs — that prove ongoing governance. Here’s why time is the most expensive variable in AI compliance.

The "Token Tax" is the unnecessary expenditure on LLM inference costs caused by inefficient prompting, redundant data processing, and lack of structure-aware document chunking. For mid-sized enterprises, this unoptimized AI architecture can lead to 40-60% revenue leakage in cloud and API costs - a "tax" that can be eliminated through intelligent, production-grade orchestration. The Hidden Drain on the Denver Tech Stack In the Denver and DTC tech corridors, we’ve moved past th

GAO report GAO-24-106214 serves as the definitive blueprint for modern federal spending audits by exposing a $195 billion discrepancy between agency records and public treasury reports. It mandates that agencies move toward more integrated, real-time data reconciliation—a process we have automated with the FedSpend platform to eliminate the "Data Lag Tax" for contractors and auditors. The Truth Bomb Heard 'Round the Beltway When the GAO released report GAO-24-106214 , it was

Under Colorado SB 24-205, any AI system that makes or is a substantial factor in making a "consequential decision"—affecting access to education, employment, financial services, healthcare, or housing—triggers strict "Duty of Reasonable Care" obligations. Failure to document this care through impact assessments and risk management systems creates an unquantifiable liability for deployers, punishable by $20,000 per violation. The End of the "Black Box" Era For years, AI has op

A "Rebuttable Presumption" of reasonable care is the primary legal shield provided by SB 24-205; AIMS secures this shield for Colorado firms by automating the NIST-aligned Risk Management System (RMS) required to shift the burden of proof in algorithmic discrimination claims. The Burden of Proof Shift In a standard discrimination claim, the burden is often on the accuser. But under the Colorado AI Act (SB 24-205) , the environment is enforcement-first. If your AI system is

Colorado SB 24-205 requires law firms to implement a Risk Management System (RMS) that includes annual impact assessments, monthly bias audits, and mandatory 90-day Attorney General notifications to establish a "rebuttable presumption" of reasonable care against algorithmic discrimination claims. The Defensibility Gap In the Denver legal community, "defensibility" is the bedrock of every strategy. But as AI integrates into the practice of law, a massive gap is opening. Most

For Colorado healthcare providers, "Reasonable Care" under SB 24-205 means implementing a Risk Management System (RMS) that actively monitors AI diagnostic and administrative tools for algorithmic discrimination—failure to do so risks $20,000 penalties per incident. The Clinical Compliance Gap In the Denver Metro healthcare corridor—from Anschutz to Cherry Creek—AI is no longer "future tech." It’s in your scheduling, your billing, and increasingly, your diagnostic support. Bu

In the legal and business world, "reasonable care" is often a gray area. But under the Colorado AI Act (SB 24-205), the state has provided a surprisingly clear roadmap for building a legal shield. It’s called the Affirmative Defense, and it hinges on a single, powerful framework: the NIST AI Risk Management Framework (AI RMF 1.0).

Most hiring managers in Denver think they are safe from the Colorado AI Act (SB24-205) because they haven't deployed a complex "AI Agent" or a humanoid robot to conduct interviews. They rely on what they consider a "neutral" tool: a spreadsheet. But here is the $10,000 question: Does that spreadsheet use a formula to rank candidates? Does it sort resumes based on keywords? Does it assign a "score" that determines who gets an interview and who gets a rejection? If the answer

Under the Colorado AI Act (SB 24-205), firms are required to exercise "reasonable care" to avoid algorithmic discrimination. In a courtroom, a PDF saying "we don't discriminate" is a weak defense. A technical log showing a Sovereign Intelligent Runtime that physically blocked a biased output is an Affirmative Defense.

The most dangerous aspect of waiting until 2026 to audit your AI systems is the Forensic Trail. Under the Colorado AI Act, "Reasonable Care" is not a checkbox; it is a documented history of governance.If the Colorado Attorney General’s office initiates an investigation on July 1, 2026, you must be able to produce: A documented history of impact assessments. Evidence of data governance measures. A record of how "consequential decisions" were monitored.

1. The Real Estate Data Crisis: Navigating the $500B Valuation Gap In the modern PropTech landscape, the delta between institutional alpha and market noise is defined by the quality of real-time property valuations. Reliable data underpins a $500 billion sector, yet the industry faces a systemic "valuation gap" caused by the fragility of traditional data acquisition. Official API channels have historically failed to provide granular valuations like Zestimates at scale, suffer