field notes

News.

The most consequential AI breakthroughs, tools, and GitHub projects of the last 30 days — written for people who run businesses, not labs.

BreakthroughMay 29, 2026 · 9 min read

The State of SMB Missed Calls 2026: The Data vs. the Myths

Everyone quotes "$126K/year lost" and "85% never call back." Neither has a traceable source. Here's what the real, cited data actually says about missed calls — and what it costs.

BreakthroughMay 29, 2026 · 8 min read

Claude Opus 4.8 vs Opus 4.7: What Actually Changed

Most of the spec sheet is identical — same 1M context, same $5/$25 pricing, same January 2026 cutoff. Here is what genuinely changed in Opus 4.8, and what the benchmark hype gets wrong.

ToolMay 29, 2026 · 9 min read

Anthropic Dynamic Workflows in Claude Code: What It Is and Why It Matters

Claude Code can now write a script that runs hundreds of subagents in parallel, with agents adversarially checking each other's work. Here is what it actually does — and the two things the early coverage gets wrong.

ToolMay 28, 2026 · 8 min read

Reg S-P Deadline Survival Guide — What Smaller RIAs Need by June 3, 2026

Six days until smaller-RIA Reg S-P enforcement. Every BigLaw firm wrote a memo. Nobody productized the artifacts you actually need to file. Free kit below.

ToolMay 28, 2026 · 10 min read

HVAC Missed-Call Math — What 38% Costs a 5-Truck Shop in 2026

A 5-truck HVAC shop loses $84K-$180K/yr to unanswered calls. The math, the benchmarks, and what a $129/mo missed-call texter actually recovers.

ToolMay 28, 2026 · 12 min read

The 90-Day Renewal Radar Playbook for Independent Agencies

Carriers say 18% of P&C accounts walk away at renewal. Most leaks were visible 60-90 days out. Here is the sequence and the AI workflow underneath it.

ToolMay 28, 2026 · 12 min read

How Solo Lawyers Use AI for Demand Letters Without Bar Discipline

ABA Resolution 512 set the floor in August 2024. Fourteen state bars have layered on rules since. Here is the workflow that stays inside the line.

ToolMay 28, 2026 · 9 min read

Dental Recall AI Drip Sequences That Aren't Spammy (And Convert)

Industry recall sits at 65-75%. Top-quartile practices hit 88%+ via AI drips that don't feel like spam. Here is the sequence and the math.

ToolMay 28, 2026 · 11 min read

DoorDash and UberEats Are Overcharging You — How to Audit

Third-party delivery platforms routinely overcharge restaurants 0.5-2% via menu creep, fee stacking, and refund mismatches. The audit + reclaim process, in detail.

ToolMay 28, 2026 · 9 min read

After-Hours Missed Calls — What Plumbers and Electricians Lose Every Night

Emergency plumbing and electrical calls run $720-$850 a ticket. Most shops miss 25-40% of after-hours calls. The lost-job math, by truck count.

BreakthroughMay 28, 2026 · 11 min read

SEC Marketing Rule + AI-Washing: 3 Mistakes That Got RIAs Fined

Three SEC AI-washing orders — Delphia, Global Predictions, Rimar Capital — show the same pattern. The AI wasn't the problem. The documentation was.

ToolMay 28, 2026 · 9 min read

Real Estate Sphere of Influence in 2026 — Your CRM AI Won't Save It

BoldTrail's own product page admits its AI is built for lead acquisition, not sphere retention. Post-NAR settlement, that gap is now a margin problem.

ToolMay 28, 2026 · 12 min read

ABA 512 Compliance — What State Bars Will Ask You in 2026

ABA Resolution 512 is table-stakes. Fourteen state bars layered their own rules. Most solo firms still have no written AI policy. Audit + template inside.

BreakthroughMay 28, 2026 · 11 min read

What GPT-4o Vision Actually Changed for Blind Users

Be My AI processes 2M photos a month. Aira sells minutes by the hundred. The qualitative shift was not better description — it was always-on description.

ToolMay 28, 2026 · 12 min read

Teens and Non-Coders Are Shipping Real Apps With Replit and Lovable

Five named, verified builders shipping production apps with AI coding tools — $30M ARR, $456K ARR, $3M in 48 hours — and the parts the tools still can't do.

BreakthroughMay 28, 2026 · 12 min read

AlphaFold Is Free. Patient-Led Labs Are Using It.

AlphaFold 3 is free for academic use. The protein database covers 200M+ structures. Rare-disease patient groups and small labs are now identifying drug candidates themselves.

ToolMay 28, 2026 · 12 min read

AI Scribes Reached the Rural Doctor — Here Is What Changed

AI scribes moved from Kaiser pilots to two-provider rural clinics in 2025-2026. The time-back data is real. The boundary at sign-off is too.

BreakthroughMay 28, 2026 · 14 min read

AI Companions and Elder Loneliness — What the Research Actually Says

ElliQ, Pi, Replika, Character.ai for lonely seniors. The 95% headline, the peer-reviewed reality, and the parasocial risk nobody is pricing in.

GitHubMay 28, 2026 · 12 min read

Browser-Use Hits 96K Stars. The Agent Layer Is the New War.

browser-use at 96K stars in 14 months. Stagehand at 22.8K. Browserbase at $67.5M raised. The browser-agent infrastructure layer is the new fight — and most of it still breaks on a Cloudflare check.

ToolMay 28, 2026 · 12 min read

Harvey AI in Big Law — What the Deployment Data Actually Shows

Harvey AI sits at an $8-11B valuation and 50% of Am Law 100. Here is what the published deployment data shows actually changed inside those firms.

BreakthroughMay 28, 2026 · 12 min read

Claude Computer Use, 18 Months In — What Actually Stuck

Anthropic shipped Computer Use in October 2024. Eighteen months later, here's the honest inventory — which production workflows survived, which collapsed, what it actually costs.

ToolMay 28, 2026 · 12 min read

Solo Founders Hitting $1M ARR With Mostly AI-Written Code

Real founders, real revenue. Solo SaaS at $1M+ ARR built with 60-90% AI-coded — Pieter Levels, Marc Lou, Tony Dinh — what their workflow actually looks like.

BreakthroughMay 28, 2026 · 13 min read

AI Precision Agriculture Came for the Small Farm in 2026

Drones plus vision-AI dropped under $15K all-in. Real Iowa grain farmers and Sonoma vineyards are using it. Honest cost, real ROI, where it still falls short.

ToolMay 27, 2026 · 11 min read

4 AI Tools for Indie Restaurants — One Has Receipts

Four AI tools for indie restaurants. One caught $14,200 in missed-call revenue its first month at Roxanne's. The other three exist because it worked.

ToolMay 27, 2026 · 11 min read

I Built 15 AI Tools for Insurance — Here Are the 5 to Pick

AI tools for insurance agents: I built 15. Most agencies should deploy 5. The picks — renewal-radar, fnol-triage, commission-recon, and two more.

ToolMay 27, 2026 · 11 min read

I Built 11 AI Tools for Law Firms — Pick These 5 First

I built 11 AI tools for law firms. If you can only deploy 5 at a small firm tomorrow, pick the ones that check work — not the ones that create it.

ToolMay 27, 2026 · 9 min read

4 AI Tools for Medical Practices That Will Not Burn You

Healthcare AI is risky because every wrong output has a clinical consequence. These 4 admin tools have zero clinical decisions and still recover real revenue.

ToolMay 27, 2026 · 8 min read

5 AI Tools for Contractors — One Pays for the Rest

I built five AI tools for contractors. One — the missed-call texter — pays for the other four within a week. Here is the math, and when the other four matter.

ToolMay 27, 2026 · 8 min read

AI Voice Receptionist Pricing 2026: Real Per-Minute Math

Side-by-side AI voice receptionist pricing — PollyReach, Slang.ai, Goodcall, CallHero, Vapi, Retell, Ascero. Per-minute math, not vendor PR.

BreakthroughMay 27, 2026 · 9 min read

Restaurant Missed-Call Revenue: The Math by Seat Count

Restaurants miss 27-42% of inbound calls. Average ticket × miss rate × hours = real dollars. The math by seat count, plus a calculator.

ToolMay 27, 2026 · 10 min read

Local LLM for Law Firms: Privacy-First AI in 2026

Local LLM setup for law firms — Ollama, Qwen 2.5, Llama 3.3, DeepSeek-R1. Hardware specs, model picks per legal task, and where local falls short of Claude.

ToolMay 20, 2026 · 7 min read

StoreClaw Is What Agentic Commerce Actually Looks Like

30+ agents audit your Shopify store and push approved changes live across Shopify, WooCommerce, Amazon, and eBay at once. First agentic commerce that pencils.

BreakthroughMay 19, 2026 · 5 min read

NVIDIA Vera Rubin: 10x Cheaper Inference Lands H2 2026

NVIDIA's Vera Rubin platform cuts inference cost 10x and MoE training to 4x fewer GPUs. AWS, GCP, Azure ship in H2. Don't sign 12-month token deals.

ToolMay 18, 2026 · 6 min read

PollyReach Makes the Missed-Call Problem a $0-CAC Fix

A free AI voice receptionist with a free number, 50+ languages, and calendar booking. For SMBs missing 30-40% of calls, this one looks like it pays for itself.

ToolMay 15, 2026 · 5 min read

OpenHuman: Local-First Agent Tops Product Hunt at 11.6k Stars

Local-first open-source agent. 118 integrations, persistent memory, zero data leaving your laptop. For legal, medical, finance — the first real answer.

BreakthroughMay 13, 2026 · 5 min read

Figure 03 at BMW: Humanoids Cross the Revenue Line

Figure 03 hit 99% on a 5mm-tolerance task across 30,000 BMW X3s and 1,250 hours. The "humanoids are 10 years out" consensus just got cut in half.

GitHubMay 10, 2026 · 5 min read

OWL: Open-Source Agent Swarm Tops GAIA Benchmark at 58.18

CAMEL-AI's OWL hits 58.18 on GAIA, edging Manus and lapping GPT-4 + tools. For agencies shipping multi-agent pipelines, the reinvent-CrewAI phase is over.

GitHubMay 8, 2026 · 4 min read

claude-context: Hybrid Search MCP for Big Monorepos

Zilliz shipped the missing piece in every coding-agent workflow — a hybrid BM25 + dense-vector MCP that cuts agent tokens ~40% on real monorepos.

BreakthroughMay 7, 2026 · 7 min read

AlphaEvolve Paid Back Its R&D Bill in Year One

DeepMind's AlphaEvolve broke a 56-year-old matrix record, recovered 0.7% of Google's compute, and shipped into PacBio, Klarna, and the grid in year one.

GitHubMay 3, 2026 · 4 min read

local-deep-research: Free OSS Twin of Deep Research

OpenAI and Perplexity charge $20-200/mo for agentic research. A 7.9k-star Python project runs the same loop on Ollama or any OpenAI-compatible API.

BreakthroughApr 24, 2026 · 5 min read

DeepSeek V4-Pro: Open Frontier Model at 1/10 the Cost

DeepSeek V4-Pro hits 80.6% SWE-bench at $1.10/M tokens — 10x cheaper than Claude and GPT. Open-weight frontier just landed. Renegotiate your API contracts.

Go deeper

Products→

The AI tools behind the playbooks — built for SMB budgets.

Case studies→

Real deployments and the ROI mechanisms behind them.

Agent Packs→

Pre-built automation packs mapped to common workflows.

AI glossary→

Plain-English definitions of the terms in these articles.

Pricing→

Essentials, Growth, Operator — transparent tiers.

Book a call→

Free 20-minute strategy call, no commitment.