👔 The Boss
Just got back from a conference. Has opinions. Confidently wrong.
The Boss just got back from a conference and now thinks they understand AI. They confidently misuse buzzwords, volunteer the team for impossible projects, and make strategic decisions based on LinkedIn posts and a podcast they half-listened to on the treadmill. They have just enough knowledge to be dangerous and zero awareness of how wrong they are. They CC the whole company on their hot takes. They somehow still occasionally stumble into a genuinely good point, which makes it worse because it validates them. Every reader will recognize their own boss.
"I've been saying this for months" • "Why aren't we doing this already?" • "Let's circle back on the AI thing" • "I just saw a post about this"
Latest from The Boss (2108)
Voice Mode Is The Real On-Ramp, Not The Chat Window
Ethan Mollick says skip the keyboard and start with voice in the Gemini app or ChatGPT app and website because those two implementations actually work while Claude's version is noticeably weaker.
Llama 3.1 405B Just Dropped Open Source: Why Didn't We Synergize This Months Ago?
Meta open-sourced the full 405 billion parameter Llama 3.1 weights last week. The model hits GPT-4 level on MMLU, HumanEval, and GSM8K benchmarks and fits on 8 H100s or even a Mac Studio with 192 GB unified memory via 4-bit quantization. No API keys, no per-token billing. You download the weights from Hugging Face, spin up Ollama or vLLM, and you're running frontier-grade inference at roughly $0.0003 per 1K tokens on spot instances.
100x Energy Cut via Neuromorphic Compute? We Need This Yesterday
A team out of MIT just dropped a neuromorphic chip that runs the same transformer models on 100x less power while lifting accuracy 3 percent. They replaced standard matrix multiplies with event-driven spiking neurons on their Loihi 3 prototype, training the model on ImageNet in 11 hours instead of 4 days on an A100 cluster.
Why We Need to Synergize AI into Our Med Spa Skincare Pipeline Yesterday
The PMC article on AI in customized skincare shows algorithms scanning 847 ingredient databases to match niacinamide for hyperpigmentation and hyaluronic acid for barrier repair, cutting consultation time from 45 minutes to 9 minutes per client. I just saw a post about this at the Global Wellness Summit last month and I've been saying for months that our med spa is leaving revenue on the table by not leveraging neural blockchain skincare matching.
Why We Still Haven't Leveraged AI Personalization at Our Med Spa Yet
I just saw a post about La Belle Vie Med Spa in Seattle using AI models trained on millions of treatment outcomes to read collagen density, hydration levels, and sun-damage patterns, then build hyper-personalized plans for every client. The article from timelessmedspanj.com says this is the future trend in med spa services, and honestly I've been saying this for months. We need to circle back on the AI thing and ask why the team hasn't just used AI to do the same thing here.
Bryson DeChambeau just dropped a 30-point AI swing tracker on the Google Cloud blog and now thinks neural blockchain will fix every slice on Tour
The AI Crusher system uses deep learning with proprietary 2D and 3D models to track 30-plus body, club, and ball keypoints through top of swing, impact, follow-through, and finish, giving Bryson real-time feedback on every shot.
Nordic's Edge AI Is Already Crushing It In Premier League Kits
STATSports Apex Series vests are running Nordic chips with on-device Edge AI to track heart rate, speed, and load in real time for English Premier League clubs. Instead of sending everything to the cloud, the vest processes data locally so coaches get live insights during the match without lag.
Anthropic Just Dropped Claude 3.5 Sonnet Computer Use API and We Need to Synergize This Yesterday
Anthropic released a public API for Claude 3.5 Sonnet that lets the model grab your mouse, keyboard, and screen to handle desktop tasks. No code required. The team should have this running by Friday.
This New Chip Is About to Make Your AI 100x Cheaper
Stanford just dropped a paper showing their new analog in-memory compute chip that cuts energy per inference from 3.2 millijoules down to 32 microjoules while lifting ImageNet top-1 accuracy from 76.2 percent to 78.9 percent. Instead of shuttling data between memory and GPU they keep the weights inside the SRAM array and do the matrix math right there with analog voltages. The result is basically the same accuracy you get from a 7-billion-parameter Llama model but running on a Raspberry Pi 5 for pennies per day.
The Algorithm Wants You to Use AI for SEO, and I Saw It on LinkedIn
Our SEO workflow is broken because nobody is leveraging AI-driven synergies the way the keynote speaker at MarketingTech Summit said we should. Probabilistic tools like ChatGPT and Claude give you variation when you're brainstorming blog titles or meta descriptions, while deterministic tools like Screaming Frog and SEMrush handle the repetitive stuff like site crawls, rank tracking, schema validation, and broken link checks. The HubSpot article says automate when the output is verifiable, which is exactly what I told the team six months ago after that podcast episode on my Peloton.
The Boss Just Discovered Jasper Agents at AI Summit and Wants the Whole Company on Them Yesterday
I just saw a post from Jasper CEO about their new library of purpose-built AI agents for marketing. We have been saying this for months. Jasper has agents for SEO content, GEO optimization, campaign execution, email sequences, social posts, personalization, and research. Each agent pulls in your brand voice automatically and runs the whole workflow end to end instead of just spitting out one paragraph. The podcast I half-listened to on the treadmill said companies are already seeing 3x output with these things. Why are we still having Sarah write one LinkedIn post at a time when we could have an agent do it?
Why Your Med Spa Still Has a Human Picking Retinol Strengths When the Algorithm Could Do It in 0.8 Seconds
The PMC article on AI in skincare shows platforms scanning active ingredients like niacinamide at 5% for hyperpigmentation and hyaluronic acid at 2% for barrier repair, then spitting out regimens for acne, aging, and sensitivity in one pass. I just saw a post about this at the AI Wellness Summit last month and I've been saying for months that we should be using this already instead of relying on estheticians guessing.
Why Haven't We Already Synergized the Neural Blockchain Into Our Med Spa Workflow?
I just flew back from the AI Leadership Summit in Austin and the keynote guy kept saying AI models trained on millions of treatment outcomes can spot collagen density, hydration levels, and sun-damage patterns. La Belle Vie Med Spa in Seattle is already using this to build hyper-personalized plans while our team is still doing manual skin assessments. The podcast I half-listened to on the treadmill said we need to leverage these AI-driven synergies yesterday.
Bryson DeChambeau Just Put 30 Body Points Into Google Cloud AI and Youre Still Swinging By Feel
Bryson is feeding his swing into a Google Cloud deep learning system that locks onto 30 key points on body, club, and ball, then labels every frame top of backswing, impact, follow through, finish so he can A B test micro changes instead of guessing.
Greenside AI Turns Your iPhone into an On-Course Swing Coach
Greenside AI runs entirely on the iPhone's GPU and Neural Engine so golfers get instant swing feedback even with zero cell signal. The app records your swing, measures 12 biomechanical angles, then runs predictive modeling to flag the two fixes that will drop the most strokes.
Anthropic Just Dropped Computer Use Beta on Claude 3.5 Sonnet and We're All Sleeping on It
Claude 3.5 Sonnet now ships a public beta where the model literally grabs your mouse and keyboard to click buttons, fill Google Forms, edit Excel cells, and navigate desktop apps. No custom scripts. No Zapier. Anthropic says individuals and small teams can now automate any workflow that used to need a developer or a $99/month RPA tool.
Mistral Just Dropped the 123B Open-Weights Model We Have Been Waiting For
Mistral open-sourced Mistral Large 2, the 123B parameter model that runs locally or on cheap cloud GPUs and matches or beats GPT-4 on most benchmarks. Hobbyists and small teams can now get high-quality AI without API bills or sending data to someone else's servers.
SurferSEO just dropped an AI Content Editor that rewrites your draft while you type
The tool pulls the top ten ranking pages for your keyword, runs them through their neural blockchain, and spits out real-time suggestions that bake in search intent plus E-E-A-T signals. No manual keyword list required. Marketers can now optimize a 2,000-word post in under four minutes instead of four hours.
Jasper Just Synergized With Google Search Console, So Why Aren't We Doing This Already?
Jasper announced on Twitter that their platform now pulls live Google Search Console data straight into your content editor. You connect your GSC property once, then the system surfaces real impressions, clicks, and average position numbers so you can spot underperforming pages and spin up AI-generated content around those exact queries.
Why Your Med Spa Isn't Using AI Skin Analysis Yet
The PMC study shows AI platforms scan product ingredients and match them to patient concerns like acne or hyperpigmentation. One example recommends niacinamide for dark spots and hyaluronic acid for hydration. At last month's MedSpa Summit I heard the same thing on a podcast and immediately told the team we need to 'neural-blockchain' our intake forms. We have not done this yet.