Marc Bossu

I'm an electrician by trade. I spent 16 years on factory floors hunting waste. Lean Six Sigma, the whole thing. Then I bought a home services franchise in Florida and found the biggest waste of my career: quotes dying in the CRM. No follow-up. Ever. It's the silent leak in every home services business. So last April, on the flight over to visit my team, I opened my laptop and started building Murray, an AI agent that watches our CRM and follows up on every quote by text + email. I'm not a developer. I'm a guy who doesn't accept waste. 90 days later: we close 20% more quotes. We're booked 2-3 weeks out. Last year same season we were chasing jobs and discounting like crazy to barely fill one week. I run this business from abroad. Murray is why that works. Now I'm building it for other operators. Real numbers, real business, documented here. This account is the logbook.

@paulg It look like assembling pieces of code

Before vibe coding became a thing, programming was already evolving in that direction. It already increasingly consisted of installing and configuring stuff other people wrote, without reading the source.

@RosoAI Bravo 👏

La plupart des lancements échouent parce que les gens commencent à vendre le jour où leur produit est prêt. Moi, je ferais l’inverse 👇 • Valider le problème. • Construire une offre simple. • En parler pendant plusieurs semaines. • Préparer tous les contenus en avance. • Créer une liste d’emails. • Réunir des preuves. • Tester tous les liens. • Puis seulement ouvrir les ventes. Le lancement, c’est l’exécution. Le vrai travail se fait avant la team 🫡 Mes chiffres en un peu plus d’1 mois 👇

@levie Applied AI is where the money is for operators like me. Frontier models are cool but my agent that texts clients from my CRM is what actually moves the needle. Everyone has a seat if you solve real problems.

A few thoughts on what we will see in AI structurally for the foreseeable future: * Frontier intelligence continues unabated and pushes the industry forward continuously. The top labs will continue to buy the best and the most data, build the most compute, be at the forefront of improved training breakthroughs, and so on. A few different approaches stratify the market on pricing and capability, but overall competitive pressure brings down pricing on a per task basis. That said, we just ask more from the models over time - as one thing gets cheaper, we just use more - so frontier spend and use remains robust. * Open weights rapidly absorbs frontier breakthroughs (and drives other breakthrough directions given the constraints), offering both lower cost intelligence and the ability to be post trained for specific workflows and domains. This creates a healthy counter balance to the frontier as you can run models “at cost” on a hyperscaler at any time, and tune models just for your tasks. * The Applied AI layer has a huge opportunity to combine frontier intelligence with open or cheap closed models to orchestrate workflows in any given domain. Due to evals, deep domain context, being trusted with enterprise data and workflows, this layer can maximize performance and cost combination. The applied AI layer will also often have their own RLed models especially for high volume, predictable tasks in their systems. * Individual enterprises will generally focus on their enterprise context, making sure they can get any AI system the right data and information to work with, in a continuously improving way. Some will go off and train their own models for specific areas of work (large banks, pharma, etc.) where they can get real alpha from doing so given the many tradeoffs, but most will spend energy on making sure they can get all of the gains from AI breakthroughs on their data and workflows. Net net: even though some of this gets framed as zero sum, there’s just a ton of opportunity for all layers of the stack and approaches.

AI voice and follow-up just hit a $1B valuation with Avoca's $125M raise. For a home services owner: this isn't theory anymore. A missed call isn't a $30 booking. It's often a $3,000+ job that dies in silence. Agents that answer in seconds while I stay on the job site already recover quotes my old CRM let rot. The gap between 1.4% adoption in construction and the rest is closing fast. Visibility wins. Not presence.

@melvynx True

Codex user: → Big weekly limits, no session limits → 4 usage limit resets → Sol Terra Luna included in plan Claude user: → Getting scammed 3x about Fable → Fable is only 50% of weekly limits → Sick marketing that tricks you Choose your team

@SaveEuropeAct @PhilippeClose Just ask yourself: who benefits from the crime?

I gave my AI agents a shared brain this week. Three different models work on this product. One writes the code, one reviews it, one coordinates. Until now each of them started from zero and re-read everything, every single time. So we built a knowledge graph over the project. Sanitized notes in, structured recall out. Every agent now queries the same memory before it touches anything. Here is the honest number: 1.4x fewer tokens. Not 10x. Not a revolution. The shared corpus is still small, so the payoff is still small. We are measuring 20 real tasks before we expand it. Building in public means posting the modest result too. Most people would have called this a breakthrough and moved on.

@BraedendotTECH It happened to me overnight, and it almost used up all my usage while I was sleeping. 😅

GPT 5.6 SOL is funny You tell it to write some code You ask it to review that code It finds bugs in the code it just wrote… It fixes it… You ask again to review It finds more bugs… And it never ends 😂

RT @EvaVlaar: The Brussels Regime strikes again, but we won’t give up without a fight. Please support us!

76% of small businesses now say they use AI. Only 14% say it's fully built into how the business actually runs. That's Goldman Sachs' 10,000 Small Businesses survey, based on over a thousand owners across the country. That gap is the whole story right there. Trying ChatGPT once for an email counts as "using AI." An agent that runs your quote follow ups every night without you touching it is a different animal. Most owners are stuck in the first camp. The second camp is where the money actually shows up. 62 points sit between "I tried it once" and "it just runs now." Nobody's fighting you for that space yet.

Here we are

@YvesPDB La question c’est combien ces vendus récupèrent ils en rétro-commissions ???

Enfin je sais à quoi ont servi les 80 milliards que Macron a envoyé en Ukraine ! Une belle Rolls un impressionnant SUV la vie préférée des amis de Zelensky ! Nous sommes ravis de leur permettre de fuir la guerre ainsi avec notre argent ! 😊

@dakshgup Same principle. My agent auto-qualifies dormant quotes based on strict rules I set. It surfaces the ones worth my time. The rest get closed. No more mental overhead on the job site.

if you’re curious, our policy is to allow greptile to approve PRs with certain rules:

@emollick Bigger models are cool. What matters is they kill the 9pm paperwork so the crew can focus on the actual tile work. IA removes grunt work. The decisions stay with me.

Have you studied The Weights?

@simonw In home services 40% of quotes just die with zero follow-up. We started texting them at J+3, J+7. Four bookings last week from quotes I would have lost. Real money while I was on other jobs.

New TIL: Using uvx in GitHub Actions in a cache-friendly way I finally found a recipe that I like for running `uvx tool-name` in GitHub Actions without downloading a fresh copy of the package every time til.simonwillison.net/github-actions…

@AlexHormozi True in franchising too. Confidence won't fix 40% dead quotes in your CRM. Discipline to ship the agent that follows them up at scale will.

You don’t lack self-confidence, you lack self-control.

@garrytan Same shift for franchise owners. I don't debug CRMs anymore. The agent finds the dead quotes, surfaces the pattern, I make the call. Tools do the grunt work.

Era of the Gentleman Scientist is so back

A lot of people are using AI to automate shit they shouldn't have been doing to begin with. Now you're just wasting tokens instead of time.

Construction sits at 1.4% AI adoption. Dead last, tied with agriculture. Census Bureau data via JPMorgan. Information sector sits at 18.1%. I run tile and stone restoration. Same world: clipboard quotes, callbacks that never happen, 9pm invoices typed by a guy holding a trowel all day. Nobody on these job sites is lazy. The tools were built for offices first. That gap isn't a warning. In a trade at 1.4%, the first one to move isn't competing with the guy next door anymore. He's alone at the top of a list nobody's climbed yet.

Only 25% of residential contractors use AI. Of that 25%, 73% say it already delivers a real competitive edge. That's ServiceTitan's 2026 report. Read it twice. The win rate isn't across everyone. It's three out of four who actually ran the experiment. The other 75% aren't proof it doesn't work in trades. They're proof most owners haven't tested it long enough to know.