No conference panels. No vendor white papers. Real deployments, real numbers, documented week by week.
The second AgentCon in Europe happened in Amsterdam in September 2026. The schedule alone told the real story: developers were on stage admitting what broke, what they misunderstood, and what they built before they understood the goal. Here is what I am bringing back to client work.
Read MoreThe call that comes eight months after an AI build ships is always the same shape: nothing broke, but the system no longer matches the business. That gap is what a managed retainer exists to close.
Read MoreA viral post this week made the point I have verified for twelve years: AI does not fix broken processes, it accelerates them. Here is the order that actually works.
Read More
The benefit of AI tooling to a business was never the argument. The real gap is friction: employees resist becoming a multi-agent employee for reasons that have nothing to do with whether it helps the company, and closing that gap means naming the friction honestly instead of re-pitching the ROI.
Read More
A client's site pulled a large volume of organic traffic that never turned into booked work. Event and engagement tracking found the real problem: wrong-keyword traffic landing users off the intended conversion path. After we fixed it, on-site conversion clicks rose from 0.2 percent to just over 1.1 percent, and the client is now booked two months out on average, on less traffic than before.
Read More
Five research items from the week of August 5, 2026.
Read More
Five research items from the week of July 29, 2026.
Read More
DRAM prices jumped 93 to 98 percent in one quarter and server memory lead times passed 40 weeks as China's CXMT opened a $4.3 billion listing to become the world's number two memory maker. Here is what the squeeze does to your hardware, cloud, and AI costs.
Read More
Five research items from the week of July 22, 2026.
Read More
Google ranks the local three pack on three factors, and only one of them is something most businesses actually manage.
Read More
A client called last quarter ready to cut their Google Ads budget because the same lead cost 22 percent more than the year before. The problem was not the budget. It was the account structure.
Read More
Five research items from the week of July 15, 2026.
Read More
I see the same pattern across the service businesses I work with: a full calendar this month, an empty one three months out. That gap is what separates the businesses that plateau from the ones that grow 45 percent a year.
Read More
A client asked this week whether their first agent was ready for production. The right question was not whether it worked. It was whether the governance around it was strong enough to let it fail safely.
Read More
A client's site looked fine on my laptop and failed the three-second test on a phone. That gap is where revenue disappears before the first headline ever loads.
Read More
A client asked me last month why I build systems instead of running a marketing agency. The answer is twenty years long and starts on a garage floor in Niagara Falls.
Read More
A client called last year because a routine plugin update had turned their checkout into a security incident. That was the moment I stopped treating platform choice as a preference and started treating it as a threshold decision.
Read More
Most of what businesses believe about LLMs comes from vendor marketing, not production. These five myths are costing teams time, money, and trust in systems that could be working better today.
Read More
Three client conversations this week all landed on the same decision: whether to keep renting AI capability or start building the operational layer around it. Here is what the research feeds said back.
Read More
Vendors sell 128K context windows. Research shows the effective usable context is a fraction of that, determined by where your data sits inside the window. The numbers from Liu et al. 2023 are specific and actionable.
Read More
I replaced our edge routing and security layer with OPNsense running on FreeBSD. Here are the kernel level network tunables that changed how AI inference traffic, large file transfers, and client pipelines behave under load.
Read More
Five items this week where the gap between research and production narrowed further. Enterprise agents are live, MCP has a structural flaw with real CVEs, and AI is compressing the time between vulnerability disclosure and active exploitation.
Read More
Tokens are the billing unit, the context window unit, and the output quality constraint in every LLM call. Understanding that triple role is the difference between a system that scales and one that produces a billing surprise at month two.
Read More
The Agentic Resource Discovery specification is not a product announcement. It is the missing infrastructure standard that makes multi-agent systems across organizations actually work, and it lands at exactly the right moment.
Read More
Five items where security moved from research to active exploitation this week. Agents are both the new attack tool and the new defense layer, and real exploits are confirming the threat model.
Read More
Most AI projects fail at the planning stage. The spec gets written before a single assumption is validated. Here is the process I actually use and why the order of operations matters more than the plan itself.
Read More
Claude Fable 5 and Mythos 5 launched June 9. OpenAI announced five new Stargate data center sites and a $500B infrastructure buildout. BBVA put 120,000 employees on ChatGPT Enterprise. Cohere shipped North Mini Code. Microsoft patched 206 flaws. And three arXiv papers documented the containment gap in deployed agentic systems. All in one week.
Read More
Five research items from the past 24 hours: agents coordinating at scale, PI-Hunter for prompt injection red-teaming, MiniMax Sparse Attention, fact-checking with RL, and GPT-5.5 in enterprise pipelines.
Read More
Four arXiv papers and one OpenAI product update all landed this week on the same gap. Agents without persistent memory spend 5,000 to 20,000 tokens reconstructing context each session. HORMA cuts that to 22% of baseline. PROJECTMEM blocks repeated failures. The substrate paper shows why a single memory score hides three separate failure modes. Ebby breaks down each finding.
Read More
Opus 4.8 ships with 1M token context, dynamic multiagent workflows, and measurably lower misaligned behavior. I have been testing it. Here is what actually changed and what it means for the businesses I work with.
Read More
I tested Gemini, Imagen, Veo, Flash, and Nano against real client workflows. Google's API stack is the most price-accessible foundation for SMB AI systems available right now. Here is what the numbers showed.
Read More
Anthropic built the most capable vulnerability-finding AI ever created, then refused to release it. Here is what that decision tells us about where cybersecurity is headed.
Read More
Sundar Pichai sat down at Google I/O and laid out a strategy that is less about hype and more about making AI work at scale. Here is what caught my attention.
Read More
Karpathy chose Anthropic over OpenAI to lead a team using Claude to accelerate Claude's own pre-training. I have been on this stack since late 2024. Here is what this hire actually signals for businesses building on AI today.
Read More
I needed to run 51,000 LLM calls to enrich a memory store. Cloud would have made the project uneconomical. Here is what local iGPU inference actually cost.
Read More
As of May 13, 2026, Google Analytics 4 automatically classifies ChatGPT, Gemini, Claude, and Perplexity referrals under a new Default Channel Group.
Read More
Intel quietly launched a $949 workstation GPU with 32GB of VRAM in March 2026. Here is what it means for private AI infrastructure and the software gap that is still real.
Read More
Small businesses are splitting into two groups: those that have integrated AI into operations and those still watching. The gap is widening faster than most expect.
Read More
Anthropic signed a compute deal with SpaceX and removed the long-context surcharge. A breakdown of what both moves actually mean for businesses building on Claude.
Read More
Google's Virgo project signals a shift toward network-level AI orchestration. What it means for agentic workflows and enterprise AI infrastructure.
Read More
AI agents are not just automating tasks. They are changing how teams are structured. What that means for businesses deploying agents in 2026.
Read More
A practical comparison of Qwen and Gemma local models for common business automation tasks including extraction, drafting, and classification.
Read More
A client had $87,000 sitting in unpaid invoices with no follow-up system. We built an automated reminder sequence tied to invoice age. Here is what happened in 120 days.
Read More
How an automated lead response system brought a home service business from 4-hour average response to under 8 minutes, and what that did to booking rates.
Read More