Enterprise AI Pilots: fail Long-Horizon agentic workflows: fail Billions and Billions spent to solve this. The solution. Enterprise software was built for humans. It needs to be built for Agents. Me explain. note on video: words are mine. not the voice. trust me, it's better that way.

if the models converge in performance, but we can tune our harness and application layer to get 35x improvements. The value creation opportunity sits with the harness and application layer, no?

when enterprise software goes headless and it will. then the best back-end wins. in headless world, agents are the users. enterprise software wasn't built for that. not a black box dot ai fixes this with Dual Representation (patent pending). a new way of writing enterprise software that is agent friendly.

Latest from notablackbox dot ai. We've just added agent access to cloud computers when using our vertical harness (built for HR). Greatly expands their capabilities and reduces latency and token costs. more here: AI Agents Are Getting Their Own Computers. HRTech Leading in Vertical Harness Development prlog.org/13176612-ai-agents…

if the enterprise software is moving headless. then the competition moves to the back-end. front-ends have multiple representations depending upon people UX back-ends also should have multiple representations to optimize for Agent UX Notablackbox.ai built 2 (filed a patent application for 1 of them) and are building a POC for a third. Best back-end wins.

Building the Machine that Builds the Machine | in Public as a quasi-hobby in early 2025, started a tiny ai lab, to solve for agentic long-horizon tasks. quickly figured out, that the challenge wasn't the models. it was "workday." we needed to rebuild "workday." so, we needed to develop a new way of building. video for deets. if, like me, you do 2x speed. volume down. unless speedcore is ur jam. turns out, not my jam. for the curious: we've got lots of agents building 24/7 every morning I get a slack message - see below (just longer) - with a summary of what they've been up to. Release Log — 2026-10-11TL;DR: A big hardening day (70 PRs across both repos) with two threads running together. Claw channels got their conversational core (an inbox every run reads, child runs per outside conversation, and a message tool over email), and the Office got leaner campaign and outputs views behind flags. Alongside that, a QA-sweep cleanup closed a long list of 500s,

How to build a $1 Trillion COMPANY in HRTech There's only one path. SERVICES revenue w/ SOFTWARE margins. How does one do this? Solve for Agents. How does one solve for agents? Build the best UX for them. But how does one solve for AG-UX? Built a tiny AI lab to figure it out.

Navier–Stoked! Much sturm und drang over AI conquering Math. AI doesn't represent impending doom to Math and obsolescence for the great mathematicians but will lead to a renaissance. How I got there: Social algos gettin crazy good.  Watched many glenn gould videos.  Rewatched the Rosalia Berghain video too many times.  then... Instagram serves up someone playing Berghain on piano in style of Gould. instagram.com/valgreenpiano/ Anyway, doom scrollin youtube, and find a Magnus Carlsen match where he did something insane the "King Walk."  Game is a few years old.  Video more recent. Carlsen did the following:  moved king forward.  moved king back.  Early in game.  Gave up castling rights.  Wasted two moves.  Opponent mad confused.  Computers go WTF. Carlsen wins.  Me delighted. I thought, for sure, AI would kill chess.  Made many of the same arguments people are making about AI taking over math. Thought, the era of kids studying games like Fischer v Spassky was over.  An arena of human ingenuity and intelligence closed off. But, chess is more popular than ever and the bar is crazy high right now.   Yes, chess was solved by AI.  But its secrets were made more accessible.  Kiddos can spend $85,000 a year to study mathematics with Tao (if they're lucky) or $500 an hour for GM tutoring if there's one about.  But the esoteric worlds of math and chess can now be explained to us as if we were 7.  And we all get to play. And if we're Carlsen, we can beat the computer or, at least, befuddle.

As frontier model capabilities converge, the opportunity moves to the application layer. When the user of the application is an Agent. Then the race is to build the best AG-UX. Gen 4 AI, is the AGI of applications. I started an AI lab to get there.

messin' around with opus 5.5 motion video. tried a banksy style mosaic on battersea power station. didn't work at all but kind of fun. open to ideas on how to fix it.

The path to building a Gen-4 AI enterprise software platform wasn't what we expected. Focus on AG-UX was the key.

building a gen 4 system of record. what is the love languages of Agents. what does an optimized AG-UX look like youtu.be/Spy3nKSLajY?si=3nma…

Thinking in Public | Hidden Tests FTW! AI is about to speedrun Goodhart's Law. The Law: "when a measure becomes a target, it ceases to be a good measure." In hiring there's much concern over the "hidden test." It typically shows up in interview etiquette. Who sits first. Is the coffee cup returned. Sometimes, a chair might be built to wobble, to see if a candidate will call it out. Understand the concern. But AI cranking out perfect CVs, perfect interview answers, perfect cover letters, hidden tests become the only possible true signal. Yelp used to be great. Once a Yelp rating became important it became useless. It was gamed. Google Maps ratings. Super useful. Now Gamed. When everything is 4.2. Nothing is 4.2. Follower counts on Instagram. A coveted KPI of influence. Gamed. Now useless. Now imagine setting a frontier model agent loose on KPIs. Backlinks are important for SEO. OK, 1000s of backlinks on their way. Holding attention on Linkedin. OK, 1000s of articles/comments inbound. Having a CV make it through AI parsers and AI screeners. OK. One, perfectly optimized CV coming up. Sales GTM motions. Optimized by AI. Become useless as ICPs bombarded. Utility of lead sourcing tools, whether sales or recruitment, wiped out. As AI systems mine and pursue targets to the point where the targets just ignore them. Pixel perfect launch videos. One prompt away. Will be, everywhere, and soon be ignored, always. Work KPIs. Gamed. Wut meanz this KPImaxxing Hiring will move to even moar hidden tests. Management will move to hidden KPIs. Foodies will develop hidden curation patterns.

Hidden Tests | Goodhart's Law in action Lots of commentary on LinkedIn about recruiters using "hidden tests" to evaluate candidates. Common example involve the etiquette of an offered coffee. Can't endorse hidden tests, but do understand them. When we built @VONQ EQO AI screening tools, Goodhart's Law was top of mind. When a measure becomes a target, it stops being a good measure. We were seeing the era of AI applications, that were being optimized for AI parsing and scoring systems. We were also seeing AI apply leading to way too many applies. We worked to introduce some more friction into the application process to address the volume of "low intent" applies. But, we also added natural language screening and interviewing into the application applies to gather more signals for scoring than just the CV. We had to build our own parser to have a common understanding of Job descriptions, CVs, interview transcripts, screening transcripts, etc. We also built an entirely new way to scoring. If we just relied upon the CV for scoring (or, as most systems do it, some sort of semantic matching), the cluster of "great candidates" gets too large to be meaningful. We work with our customers to help solve this. So that they don't need to keep track of who offers to return the coffee cup to the kitchen

Yes. One can vibe-code a Workday in their evenings and weekends. The catch, 500+ days and counting of daily updates that look like this. Agents cook. We review. Repeat.

Vertical Harness + Enterprise App to reinvent enterprise software. not a black box dot ai

stream of claude opus 5.5 videos in my timeline. looked good. but, me dubious. spent 3 minutes on one while eating leftovers. now wish I spent 30 minutes. Or, at least 15.