Insights
Field notes on governed software.
How operating companies put AI to work without handing over the keys — the thinking behind propose, never commit, plus what we’re building and what we’re learning in the field.
Why governed AI beats autonomous agents
Fully autonomous agents demo well and operate badly. For companies with real liability, the winning design isn't more autonomy — it's AI that proposes, with a human on every action and a trail behind it.
Propose, never commit: why your software should stage, not act
“Propose, never commit” is the rule under everything we build. Here's what it actually means in the product — the four steps every action moves through, and why the audit trail is a feature, not paperwork.
The Gatekeeper Script Is a Symptom, Not the Problem
Brokers keep trading scripts for getting past the "reason for calling" screener. A crisp line helps once — but the gatekeeper isn't beaten by a phrase. It's beaten by not being a stranger: a prospecting record where every attempt, note, and follow-up is captured, so call two opens with context instead of a cold pitch. Here's how to build that record without more admin.
The Startup "Fitness Stages" Lesson, Read for a Growing CRE Shop
A construction-fintech CEO's "5 fitness stages" framework was written for founders — but it answers a question every growing CRE shop is already asking: as headcount and deal volume climb, what do I let my systems do on their own, and what still needs a human hand on it? We translate startup growth stages into a staging discipline for the tools an owner-operator buys.
A $3.5B Valuation Is an Operations Story, Not a Building Story
Brookfield is buying into the $3.5 billion Hudson Square portfolio as tech tenants push rents up roughly 20%. But a leasing surge rewards the owner whose process can absorb it — capturing tours, broker updates, and lease expirations in one place — not the one with the best building. Here's what to fix before demand tests your pipeline.
Undervalued CRE Sectors: The Edge Isn't the Tip, It's Your Data
Everyone wants the overlooked CRE sector — IOS, small-bay flex, secondary medical office, tertiary self-storage. But the sector is only half the answer. The durable edge is being organized enough to source, underwrite, and close a niche deal before the big funds crowd in. That edge comes from a clean system of record, not a hot tip. Here's what operators keep naming, and what actually lets you move on it first.
The Broker Privacy Demand Letters Are a Governance Problem, Not a Website Problem
Serial plaintiffs are scanning brokerage websites for tracking pixels and session recorders, then sending California privacy demand letters. A legal defense fund helps after the fact — but the real fix is treating every outbound data flow the way you'd treat any consequential action: stage it, log it, let a human approve it. Here's what triggers these claims and how to close the exposure this week.
Your Team Is Already Using AI. You Just Can't See It Yet.
Jakob Nielsen's June 2026 UX Roundup documents explosive agentic-AI use among non-developers — and a gap between what people say and what they do. For CRE operators, that means your leasing team is already drafting emails, LOIs, and rent-roll summaries with AI you can't see. The fix isn't a ban. It's governance: stage the action, let a human approve, keep the record.
How to Migrate Off AccuLynx Without Losing the Follow-Ups
AccuLynx is a roofing-contractor CRM, so migrating off it is more than a CSV export — it's a move to a system that fits leasing pipelines, rent rolls, and broker updates. This guide covers what to export, what quietly breaks (the open follow-ups), and how to stage the import so a human approves every batch instead of trusting a blind bulk import.
The Frameworks That Outlast You: What CRE Operators Should Take From the CXO Interviews
A recent round of CXO interviews keeps landing on the same point: the frameworks that outlast their creators are the ones written down where others can follow them. For CRE operators, the durable asset isn't the principal's instinct — it's the leasing cadence and renewal rules captured in a system your firm, and its software, can actually act on. Here's how to make that framework survive.
The Appeal Nobody Files: What CRE Can Learn From the Denial-Packet Pattern
Fewer than 1% of denied claims get appealed — not from apathy, but friction. The same unfiled-appeal problem sits in every CRE portfolio: property tax protests, CAM disputes, lease-clause overcharges. Here's how the "denial-packet" pattern — structure the pile, cite everything, stop before sending — turns a document pile into a reviewable case, and why the stop matters more than the draft.
What Legal Tech's AI Fight Tells CRE Operators to Ask Instead
Legal tech spent Q4 2025 arguing over general-purpose vs. legal-specific AI. CRE operators are about to have the identical fight — and it's the wrong one. The model isn't the deciding factor. Whether the tool knows your deals, and whether it proposes or acts on its own, matters far more. Here's the question to ask instead, and how a governed leasing workflow actually handles it.
Pass-Through Property Taxes: What They Mean for a New Owner
A property listed with "pass-through property taxes" isn't handing a new owner free rent — it's a recovery structure that still runs the tax liability through you. This guide explains what pass-throughs mean, how reassessment on sale changes the math, the difference between NNN and base-year clauses, and where the recovery terms should actually live so the annual reconciliation never quietly gets skipped.
Building an After-Tax CRE Partnership Model You Can Actually Trace
Most CRE acquisition models stop at pre-tax levered cash flow. Extending to after-tax investor returns means chaining four dependent calculations — outside basis, suspended passive losses, depreciation recapture, and long-term capital gain — in the right order. This guide walks the flow from acquisition to disposition and argues the real risk isn't a missing formula. It's a silent wrong one. Build in Excel; govern the handoff.
The "AI Utility Belt" Is the Wrong Model for a CRE Firm
The "AI utility belt" — a kit of chat tools for drafting, summarizing, and prospecting — helps an individual move faster but leaves a CRE firm with drafts nobody approved and numbers nobody checked. Here's why an operating company needs one governed desk instead of a scatter of point tools, and where the belt is still fine.
Non-Responsive Tenants and Missed Deadlines Are a Tracking Problem, Not Just a Manager Problem
A Houston owner asked online for a property manager who can handle non-responsive tenants and evictions. The better question is what tracks the follow-ups, notices, and lease dates that decide those cases. Whether you hire out or self-manage a small rental, the fix is a governed system of record that stages the next action and keeps a defensible trail — not just a better vendor.
An AI Agent for CRE Just Shipped. The Real Question Is Who Approves.
A CRE newsletter just shipped its own AI agent and four ways to use ChatGPT's o3 model. The four uses are genuinely useful — because a human reads every output. An autonomous agent with access to your rent roll and outbound is a different animal. Here's how CRE operators should separate a governed assistant from an agent with keys to the system of record, and the four questions to ask first.
What a 25-Year AI Veteran Gets Right About CRE Software
Adi Shavit spent 25 years building AI systems, and his lesson for startups is the same one stalling CRE pilots: start from the real problem, not the technology. We translate that into operator terms — name the broken artifact, stage the fix, and let a human approve it — so your next software bet retires a workaround instead of chasing a demo.
The AI Expectations Gap Is Real. CRE Operators Should Close It the Boring Way.
Legal tech just named the "striking divergence in AI expectations." CRE operators are living the same gap: leaders expect transformation, the front line sees little change. The fix isn't more enthusiasm or more autonomy — it's governed, narrow workflows a human approves. Here's how to pick the first one and close the gap the boring, defensible way.
The Token Limit Is Teaching You the Right Lesson About AI Context
Developers cut wasted context to stop hitting Claude Code's usage limits. The trick isn't a billing hack — it's a lesson operators need. Feed AI narrow, layered context instead of a giant memory dump, and it guesses less, costs less, and produces proposals a human can actually review. Here's how the token limit points straight at governance.
Capability Outran Usability. Governance Is How CRE Closes the Gap.
Halfway through 2026, AI models are moving faster than the interfaces built around them. For commercial real estate operators, that gap explains why impressive demos keep stalling in production. The answer isn't a bigger model — it's governance: software that proposes, a human who approves, and an audit trail that remembers. Here's what that looks like against a rent roll and a leasing tracker.
AI Search Won't Cite Your Homepage. It Cites Your Answers.
Homepage advice misses the point for CRE firms chasing AI-search referrals. Answer engines cite the page that answers a specific operator question — lease expirations, rent rolls, broker updates — not the page that describes your company. Here's how to build citable answer pages, why "propose, never commit" applies to published claims, and which CRE questions prospects actually put to ChatGPT and Perplexity.
How to Underwrite Sale-Leaseback Proceeds: Model the Use, Then Verify It
Sale-leaseback proceeds sit awkwardly in an SLB credit: underwrite off history alone and you ignore a real event, but credit the full pro forma and you underwrite a promise. This guide walks the sequence net-lease and private-credit underwriters use — anchor on history, model the intended use of proceeds, then make that use a closing condition and verify it before you fund.
The Google Rank #1 Trap: Why Answer Engines Decide Who Gets Cited
Ranking #1 on Google no longer guarantees the click — answer engines resolve the question before anyone visits. This piece explains the rank #1 trap, why AI-referred traffic often converts better, and how operators can audit their buyer questions against ChatGPT, Perplexity, and Google's AI Overviews to become the source the machine actually cites.
The Demo Reel Won't Tell You Which Agent to Trust
Weekly "show off your agent" threads are full of clever demos — and demos are optimized for the wrong thing. For operators, the question isn't what an agent can generate but what it can be trusted to do when no one's watching. A practical checklist for judging agents by their audit trail, not their highlight reel.
What Actually Makes an Accounting Firm Worth More in 2026
Private equity is repricing accounting firms, and "AI-native" is showing up as a valuation premium. But autonomous automation is a black box a buyer discounts. Governed automation — software drafts, a human approves, everything lands on an audit trail — is what survives diligence. Here's how buyers actually value firms in 2026, and why the audit trail is the asset.
A CRE Newsletter Just Introduced Its Own AI Agent. Here's What Operators Should Actually Ask.
A CRE newsletter just introduced a named AI agent with a friendly hello. That's marketing. Before you let any agent near your pipeline, the real questions are about permissions: what it reads, what it sends on its own, who approves, and whether every action lands on an audit trail. Here's the checklist operators should run — and why "propose, never commit" outlasts any persona.
What an "AI Operating System" Actually Means for an Operating Company
"AI operating system" is a fashionable phrase with a fuzzy definition. For an operating company, the useful meaning is narrow: a governed layer that reads your business context, proposes actions, and waits for a human to approve — with everything on an audit trail. Here's what the term should mean, what an AI OS should include, and why acting on its own is the part to avoid.
How to Track Lease Expirations Across Your CRE Portfolio
Lease expirations get missed because the spreadsheet is a snapshot nobody re-dates — not because owners forget their leases end. This guide covers critical-date tracking, why rent rolls drift, and a system that alerts you ahead of every notice deadline and stages tenant follow-ups you approve. Written for CRE owner-operators managing renewals across a portfolio.
The Context Layer: Why AI Only Works When It Knows Your Business
Everyone can rent the same AI model. Nobody can rent your firm's records, rules, and history. That's the context layer — the structured, governed knowledge that turns a generic model into a system that knows your business. Here's what it holds, why bigger models don't replace it, and how to start narrow without handing over the keys.
Why AI Search Can't Cite a Site It Can't Read
AI answer engines can't cite a page they can't read — and won't cite one they can't easily quote. This guide separates the durable GEO fixes operators control (crawler access, answer-first first paragraphs, extractable chunks) from the viral "150% in 30 days" hype, and shows how to stage each change, verify the citation, and keep the audit trail.
Agents and Accuracy Guarantees Are Coming for Your Accounting Firm. Read the Fine Print.
Vendors are shipping AI agents and "tax accuracy guarantees" to accounting firms. Both ask you to trust software that acts on its own. This is the operator's read: a guarantee pays out after a mistake — it doesn't stop one. The safer default is governed automation, where the software drafts and a human approves before anything is committed under your license.
OpenClaw for CRE Prospecting: Gather Freely, Send Carefully
Scraping tools like OpenClaw make CRE prospecting data cheap to gather — but gathering was never the hard part. The risk lives in outreach. This guide explains why the send button, not the scrape, is where legal and reputational exposure sits, and how a propose-never-commit workflow keeps a human accountable for every message that leaves the building.
The AI Questions CRE Operators Should Actually Be Asking in 2026
The year-end AI roundups for commercial real estate keep asking what the technology can do. That's the wrong lead. In 2026, capability is mostly solved — governance isn't. This piece reframes the questions CRE operators should actually ask: not what AI can do, but what it's allowed to do, who approves it, and where every action gets logged.
Can you trust ChatGPT with your rent roll?
Someone on your team has already pasted lease data into a chatbot. The real exposure isn't the model reading your rent roll — it's ungoverned tools writing to your records. Where the risk actually lives, the five questions to ask any AI vendor, and the read→propose→approve→audit pattern that makes adoption safe.
How an industrial developer replaced its leasing tracker in a weekend
A Central Florida industrial developer ran a multi-park portfolio on an 11MB Excel tracker. In one weekend the tracker seeded a governed system of record — bay map, computed rollups, attributed edits — and only then did AI arrive: broker packets and texted updates parsed into one propose-never-commit review queue.
Claude for commercial real estate teams: what it's actually good at
We run Claude in production for CRE work daily. Four uses hold up: lease/document reading, structuring messy broker updates into data, drafting, and portfolio Q&A — each with one specific guardrail. Plus the rule of thumb for when to use chat directly versus inside a governed system.
Why 92% of CRE firms pilot AI — and so few reach their goals
Colliers reports ~92% of CRE firms have piloted AI while only ~5% reach their goals — the "AI productivity gap." The gap isn't model quality: pilots bolt AI beside the workflow instead of into the system of record. What the successful minority does differently, and a concrete this-quarter plan for mid-size firms.
AI lease abstraction: what it reads right — and where it needs a human
AI turns lease abstraction from a paralegal week into minutes — but the trustworthy version is a workflow, not a chatbot. What models reliably extract (dates, rent, named clauses), the three places they consistently miss, and the four-step review workflow that makes the output safe to run a portfolio on.
When does a leasing spreadsheet stop being enough?
Leasing trackers don't fail at math — they fail at concurrency and memory. Five concrete signs the spreadsheet is costing you deals (version drift, silent collisions, expirations by memory), why follow-up breaks before data does, and what a one-weekend migration off Excel actually looks like.
What if brokers could just text the CRM?
Deal flow gets reported by text — and usually dies there. Text-to-CRM capture parses each message into a structured, staged suggestion a person reviews and promotes: capture without re-keying, speed without surrendering write access. How it works, why it doesn't corrupt the record, and what it does to the leasing meeting.
Do CRE teams need a CRM — or a system of record?
Lists of "best CRE CRMs" assume your problem is contacts. For owners, developers, and leasing teams, the real gap is a system of record — properties, suites, and leases as the spine, with contacts attached. Here's how to tell which one you actually need, and when purpose-built beats configurable.
Using AI to Find Tenants for Commercial Real Estate — What It Actually Does
"Use AI to find tenants" hides several different jobs — prospecting, outreach, and qualification. Software is strong at the first two and should stay out of the third. This guide maps the CRE leasing workflow, shows which data you already hold, and explains why the safe version has the desk draft and a human send every message.
AI Listing Photos Are Free. Being Wrong About the Property Is Not.
AI listing imagery is effectively free now, so the cost has moved from making the picture to standing behind what it claims. This guide covers which AI-generated and enhanced images a CRE firm can defend, when disclosure is required under NAR and FTC standards, and the propose-never-commit workflow that puts a named reviewer on every published image.
Stop Babysitting Six AIs: Fix the Handoff, Not the Model
If you feel like you're babysitting six AI tools to finish one task, the problem isn't the models — it's the handoff. Work stalls at the seams where context, state, and receipts get dropped. Here's why a shared, readable task record with a human approval at each step beats stitching outputs together by hand.
A Secure Proxy Isn't Governance: What "100+ Integrations" Really Asks of You
An open-source agent tool made secure credentials its headline feature: keys stay on the host, and you can cut the agent off from the internet entirely. Good engineering — but it answers the wrong question. Securing the connection isn't the same as governing the decision. Here's what "100+ integrations" actually asks of an operator, and the questions to ask before you hand over the keys.
61 New Accounting AI Agents Later, the Real Question Is Who Approves
One accounting-AI round-up counted 61 updates in a single cycle, most of them agents built to act on the ledger. For a firm owner, the count isn't the story — the sign-off is. Here's why governed automation, where the desk drafts and a human approves, beats autonomous agents for anything that touches books, payroll, or treasury.
Why Governed AI Beats Autonomous Agents
Autonomous agents optimize to remove the human — the one control that catches a wrong invoice or a bad email before it ships. Governed AI does the same drafting work but stops one step short: it proposes, a person approves, and the action lands on an audit trail. Here's why that pattern wins in production, and the four questions to ask before you turn any tool loose.
When Generation Is Free, Judgment Is the Job
Generative models made producing a first draft nearly free — but production was never the bottleneck in professional work. The costly, risky part is deciding, checking, and sending. Here's why the firms that win the next wave won't generate the most; they'll govern review best. A concrete case for "propose, never commit," backed by named studies on where knowledge-work time goes and how often AI errors reach clients.
SaaS Diligence: The LOI Is Where the Real Work Starts
Internet narratives make buying a small SaaS sound fast and clean. It isn't. The signed LOI isn't the finish line — it's where you start proving the seller's story survives the data. Here's the diligence checklist that matters, and why treating every claim as a proposition to verify keeps first-time buyers from paying for a story instead of a business.
AI CRE Videos Are Cheap. Wrong Numbers on Camera Are Not.
AI video tools make a CRE listing walkthrough in minutes for a few dollars, so production is a solved problem. The real decision is governance: who verifies availability, price, and square footage before a synthetic presenter states them on camera. Cheap production raises the cost of a public error — here is the propose-never-commit workflow that keeps the speed and the check.
Should You Build Your Own AI Memory? For an Operating Company, Not Like That
"Build your own AI memory" is fine for one person and a trap for a firm. A private memory blob solves recall but fails on the three things that matter: it walks out the door when someone quits, it drifts as facts change, and no one can audit it. Here's the governed, layered alternative operators should build instead.
The AI Agents That Actually Ship: Why Governed Beats Autonomous in Production
Operators keep asking whether anyone has deployed an AI agent that actually saves manual work. The honest answer: the ones that last are governed. An autonomous agent that acts on its own multiplies your exposure to hallucination, token cost, and memory drift. A governed agent drafts the work and waits for a human sign-off — which is exactly why it stays in production.
Why ChatGPT Keeps Recommending the Same Law Firms
Ask ChatGPT for local law firms and the same names keep surfacing. That repetition isn't random or bought — it reflects which firms an answer engine can identify with confidence. Here's what the pattern actually rewards, how answer engine optimization differs from SEO, and how to make your firm legible without handing a regulated practice's public claims to unsupervised automation.
Your AI-Search Dashboard Counts Mentions. It Should Diagnose Them.
AI-search visibility dashboards tell you whether an engine mentioned your firm — not why it cited a competitor instead. This piece argues the useful tool diagnoses the mechanism behind each citation, then stages fixes for a human to approve. A diagnosis-first, governed approach to Answer Engine Optimization for operators who want cause, not another vanity score.
Does Your Product Page Answer the Buyer's Question in the First 100 Words?
Most product pages hide what they sell behind a tagline and a hero image. In 2026, buyers skim and answer engines quote — so the page that states its answer in the first 100 words wins. Here's how to test your opening block and rewrite it to answer the buyer's real question: what is it, who's it for, and why choose it.
The 4 Questions to Ask Before You Give an AI Tool Access to Your Business
New AI tools ship weekly, and each one asks for access to your files, inbox, and messages. Owners approve these grants like a normal plugin — but a tool that can read your data and act on it is a standing commitment. Here's the four-question test to run before you connect anything, and how to roll it out in stages.
The Five-Line Homepage Test: Category, ICP, Pain, Proof, Alternative
A homepage has one job: let a stranger repeat back what you are, who it's for, what it fixes, why to believe you, and what you replace. Most operator sites read fluent and say nothing. Here's the five-line test — category, ICP, pain, proof, alternative — with what "good" looks like and a ten-minute way to run it yourself.
Can AI Search Understand Your Homepage?
Answer engines don't admire your homepage — they parse it for entities, claims, and structured answers. A homepage built from adjectives gives them nothing to extract, so it goes uncited. This piece shows operators how to write a page that both a buyer and an AI search engine can actually understand, and why plain, sourced language beats slogans.
Layered Context Beats Big Memory: How Operators Keep AI Reliable
Real AI workflows are converging on a quiet conclusion: bigger memory isn't the answer. Operators get reliable results by layering context into small, named files — stable preferences, project facts, task context — and deleting what's stale. Here's the file layout that works, why handoffs beat marathon sessions, and how layered context is the same instinct as governed, inspectable software.
When Perplexity Recommends Your Competitor, Read the Citation
When an AI engine recommends a competitor, the citation tells you why: they published a quotable, source-backed answer and you didn't. This piece reframes the recommendation as a measurable coverage gap — and lays out a plain loop to find which questions you failed to answer, and how to close them without letting a bot publish in your name.
Track Every Lead and Voice Note in One Place — Without Losing the Details
Tracking incoming leads and capturing voice notes on the go are the same problem: getting scattered inputs into one trustworthy place. Here's a simple capture-to-propose setup where a forwarded email, text, or voice memo becomes a draft entry — and nothing commits until you confirm. Fast capture, no lost details, one system of record.
MCP is standardizing. The governance question isn't.
MCP standardized fast in early 2026: harnesses consolidated, stores opened, and the first official UI extension merged. The plumbing is settled — the governance isn't. As value shifts from agents to servers, the real question for operating companies is what an agent's MCP tools are allowed to do. Here's why write access, not deployment, is the risk to design around.
Agents Are Slow Because Acting Is the Hard Part
Mark Zuckerberg says AI agent development is going slower than expected. That's not a scaling delay — it's a design lesson. Acting is harder than proposing, because action means irreversibility, authority, and accountability. This piece argues for governed AI that stages proposals for human commitment, and explains why that model ships today while full autonomy keeps slipping.
The AI Evangelist Is a Symptom, Not a Strategy
AI evangelism is loud right now for a reason: when software acts on its own and can't be verified, all that's left to sell is belief. We take an Ask HN complaint about insufferable "AI thought leaders" and trace the noise back to autonomy — and argue that governed software, which proposes instead of commits, is what actually quiets the hype.
Simple Automations for a CRE Firm (That Won't Blow Up in Your Face)
The automations worth starting with at a CRE firm aren't the impressive ones — they're the boring, repetitive handoffs: intake, follow-up drafts, deadline reminders, document routing. Start where the rules are clear and a mistake is cheap, keep a person in the approval seat, and put every action on an audit trail. A practical guide to what to automate first, and what to leave alone.
The AI Questions CRE Operators Actually Ask in 2026
The 2026 "AI use case" lists for commercial real estate mostly show tasks getting automated and stop there. Operators are asking a sharper question: what will the software do without asking me? This guide reframes the top CRE use cases — abstraction, rent rolls, county records, outbound — around control, and explains why "propose, never commit" is the right default.
The Top AI-for-CRE Questions of 2026 — Answered
The AI questions CRE operators actually ask in 2026 aren't about models — they're about records, intake, outbound, and CRMs. Here are the top ones, answered plainly, with a single test underneath all of them: does the software propose and wait for you, or does it commit on its own? Governance, not novelty, is the real evaluation.
Governed Outbound: Let the Desk Draft, Let a Human Send
Autonomous senders promise booked meetings and deliver public mistakes at scale. Governed outbound flips the pattern: the system researches, drafts, and queues every message on an audit trail — and a human approves the send. Here's what to automate, what to keep human, and why staging every send still scales for firms that sell by relationship.
What the $200K-MRR "cannot fail" playbook teaches operators about de-risking software
A November 2025 Starter Story interview breaks down how one founder built a $200K-MRR portfolio of small SaaS apps by refusing to guess whether anyone wants the product. We pull out the operator's version of the lesson: de-risking a business and governing your software are the same instinct — pick proven bets, keep early cash disciplined, and keep a human deciding what ships.
CRM or Project Tracker? For CRE Firms, It's One System of Record
CRE firms don't need a CRM and a separate project tracker — they need one governed system of record that both read from. This guide explains why the split causes dropped follow-ups, how to build role-based dashboards as filtered views, and where "propose, never commit" keeps the record honest. Practical setup steps included, written for owners and operators, not IT.
AI Integration for CRE Firms: Propose, Don't Commit
"AI integration" for commercial real estate is usually sold as autonomy — agents that send, file, and act. For CRE, where a single wrong filing is real exposure, the better default is the opposite: AI that reads your county records and rent rolls, drafts the work, and stages every outbound action for human sign-off. Here's where to put it, and how to keep it governed as it scales.
A Quality Standard an Agent Can Read Is Governance, Not a Guarantee
QUALITY.md lets a project declare its quality bar in a file that both people and coding agents can read. That's genuinely useful — but only if the agent proposes against the standard instead of self-approving. Here's why a machine-readable quality file is a governance artifact, not a quality guarantee, and how to adopt it without laundering unreviewed work as compliant.
Why governed AI beats autonomous agents
Fully autonomous agents demo well and operate badly. For companies with real liability, the winning design isn't more autonomy — it's AI that proposes, with a human on every action and a trail behind it.
Propose, never commit: why your software should stage, not act
“Propose, never commit” is the rule under everything we build. Here's what it actually means in the product — the four steps every action moves through, and why the audit trail is a feature, not paperwork.
How one developer built entitlement intelligence across six jurisdictions
An industrial real estate developer operating across the Southeast needed to know — without a research analyst per county — when zoning and entitlement activity actually mattered to them. So we built it, configured to their footprint.
The data CRE firms are sitting on: county records + AI
Public records hold most of what a real estate operator needs to know — and almost none of it is usable as-is. A practical look at turning agendas, filings, and minutes into a sourced, ranked feed you'd actually read.
Reading is the slow way to see it.
Thirty minutes with your own workflow beats any article.