LLM Future: Transformers Evolve, Inference Scaling Takes Spotlight
Had a fun chat with @mattturck the other day where we talked about a bunch of interesting LLM stuff... Basically everything from the future of the transformer architecture to inference-time scaling as a recent MVP of LLM performance:
Arcee Trinity Large Merges MoE, Gated
It's been a while since I did an LLM architecture post. Just stumbled upon the Arcee AI Trinity Large release + technical report released yesterday and couldn't resist: - 400B param MoE (13B active params) - Base model performance similar to GLM...
Apple’s iPhone Voice-to-Text Still Mishears “How AI”
I just used voice-to-text on my iPhone for a text message in which i spoke, “how AI.” My iPhone typed “Howie.” I feel like that about sums up Apple’s current AI capabilities on the iPhone.
AI Training Could Double Worker Productivity, Study Finds
Does AI training double worker productivity? What's your reaction to this study from Microsoft? Follow for more.
B2B Marketers Shift From Paid Ads to Events, Partnerships
The great paid ads pullback of 2026 is here. We surveyed 50 mid-market B2B marketing leaders with Wynter. The data is violent. 72% are actively slashing budgets for paid ads, content syndication, or agencies. This isn't a "hot take." It is a capital...
Robotics Teams Lose Time Rebuilding Infrastructure, Not Algorithms
🎙️ Stephen James is the founder and CEO of Neuracore , and Assistant Professor of Robot Learning at Imperial College London: In this episode, Stephen shares his path from growing up in Wales to spending a decade at Imperial, a...
Robotics Demos Lag Far Behind Real-World Production
there is massive arb between robotics demos you see on twitter and what very few companies have in large-scale production today and no, the gap isn't in favor of the twitter demos.
QMD Launches Native Query Extender, Boosting Expansion Performance
QMD now has its own natively trained query extender model https://t.co/b89CmKIknV Seems to do much better at query expansion in my tests. try it out.
Rolv's AI Speed Breakthrough Shows Promising Results
Months ago I shared Rolv with you and he shared that he had a major performance breakthrough for running AIs. His breakthrough is proving itself. Will share the final results when it's up, but his math is working out to...
Brayden's Architecture Beats ChatGPT for Molty Reports
The second "Ultimate Molty Report." Thanks @blevlabs. If it weren't for Brayden's cognitive architecture these reports wouldn't be nearly as good as if I used, say, Grok or ChatGPT to write them.
New AI Releases and Viral Moltbot Report Incoming
Tomorrow morning @higgsfield_ai and @AnthropicAI will drop something crazy. And I'm about to drop another @moltbot report. My first one got viewed hundreds of thousands of times.
AI Agents Improve Instantly with Each Compute Boost
We've had a pretty wild new AI Agent feature that we've almost released twice, but then right as we get close, a new technical breakthrough makes the agent substantially better than before. The pace of new best practices emerging in...
VCs Can't Fund Ex‑OpenAI Founders Indefinitely
“VCs can’t just keep giving OpenAI employees billions of dollars to start a new company forever” VCs: https://t.co/lVyo5wnDoI
Scaling AI Developer Infrastructure: Lessons From Meta’s Internal Agent
Office Hours: Building Developer Infrastructure at Scale with Jim Everingham In the world of AI, how do the very best companies build their developer infrastructure? To answer that question : On Tuesday, February 24, I’ll be hosting an in-person fireside...
Reviving BabyAGI Amid Surge of Build Energy
it’s been so fun watching the build energy this past week, i had to start rebuilding babyagi again last night :)
AI Brand Rankings Unreliable; Track Mention Frequency Instead
Ranking position in AI responses is meaningless. Across repeated runs of the same prompt, LLMs almost never return the same list of brands in the same order. In many cases, they don’t return the same list at all. New research...
ChatGPT Could Power Universal Price Comparisons via Shopify Catalog
I saw a Q&A that mentioned @Shopify's Catalog MCP provides a "List of shops selling the product with specific prices" along with a Universal Product ID. Does that mean a shopper could use chatgpt for easy price comparison shopping across...
Prioritizing Guided Onboarding for Complex Data Platform
I spent 2025 building out the best product I could for brand and category tracking, and my current next product focus is making sure that I surround it with a strong onboarding and training experience. The double-edged sword of adding...
Intercom's AI Pivot Redefines SaaS Economics
Intercom has pivoted from traditional SaaS to AI better than any company I can think of. I spoke to the CFO behind the customer support juggernaut, and the outcome based product Finn. Dan Griggs is mastering the new economics of...
Nori AI: Simple, All‑In‑One Family Planner
Clawdbot, now called @moltbot, is hot, but sometimes you just want a simple family planner, and don’t want to buy a new Mac Mini and nerd out for the weekend just to get a family planning system running. Nori is...
AI Success Shifts: From Language Masters to System Designers
We’re seeing the real-time shift in who wins with AI. Here’s my winners list. ↓ ↓ ↓ 2023 - English majors. They understood that the model is a language machine before it is anything else. Their expanded vocabulary and precise...
AI Ended Dropshipping; Human Strategy Wins 2026
AI just Killed Dropshipping... Here's What's Replacing it Today Watch here 👉 https://t.co/ZExaWfvw02 AI has turned dropshipping into a race of copycats. In this video, I explain how automation is pushing beginners into competing on the same products, killing stores before...
Clean Customer Data Powers Personalized LLM Responses
Tealium gives LLMs what they usually lack: real customer context. Things like cross-channel behavior, lifecycle stage, affinities, and history flow into prompts and workflows, so AI responses stop feeling generic and start feeling more like they’re written for one person....
LinkedIn Flags AI‑generated Images with New Content Label
I made this image on ChatGPT and added it to LinkedIn. Never seen this "Content credentials" label appear before - looks like LinkedIn is telling people when images are 100% AI-generated. 🤔 (Sorry if not new, it's new to me!) https://t.co/ZXENZmGKox
AI‑Powered Concept Testing Eliminates Product Validation Bottleneck
The bottleneck in product development is moving towards knowing what to build. Ignoring it is going to lead to bloated products. Pre-AI it was already common to skip validation. But as execution speeds up, and you need to feed the...
ChatGPT's Recommendations Vary Widely, Challenging Brand Tracking
NEW research: https://t.co/yE1LogIFHO If you give ChatGPT the same request for product recommendations 100X, will you ever get the same list twice? And what does that answer mean for folks who try to track their brand presence in AI tools? Finally, some real...
Early AI Adoption Gives Competitive Edge in Customer Service
The 2026 Customer Service Trends Report is out. Every year we interview over 2,000 CS leaders to discuss what's happening in AI & Customer Service, and publish a full 50 page report of everything we've learned. 4 Key ideas for...
Open‑Source AI Gains Power, Raises Massive Security Risks
On one hand we should expect many open source models to get great at computer use because of clawdbot proving demand On the other hand random free oss models controlling millions of computers sounds like a nightmare
Build Reusable Anthropic Agent Skills for Any Workflow
Important new course: Agent Skills with Anthropic, built with Anthropic and taught by Elie Schoppik ! Skills are constructed as folders of instructions that equip agents with on-demand knowledge and workflows. This short course teaches you how to create them...
Open‑Source Dev Kit Enables AI‑Native Robotics Integration
An Open Source Dev Kit for AI-native Robotics. [📍github below ] This repo uses LeRobot's plugin conventions to be automatically detected by a LeRobot installation in the same Python environment. https://lnkd.in/gc5tmifi —— Weekly robotics and AI insights. Subscribe free: scalingdeep.tech
M5 Ultra Promises Double AI Speed, Dual Studio Dream
M5 Ultra should be coming in a few months. Rumors say up to 100% faster AI compute than the M3 Ultra. A dual M5 Ultra Mac Studio setup for 2026’s best local models will be quite the home AI setup.
AI Revolutionizes Enterprise Sales and Organizational Structure
New @ThePeelPod with @JacquelineSYC19 How AI is changing data, getting enterprise customers from cold emails during YC, building an AI BDR in-house, why most functions at @artie_labs report to the CTO (not CEO), and what it's like raising a Series...
Mercedes‑Teams Partnership Clashes on Brand Coherence
Mercedes is integrating Microsoft Teams into the 2026 all-new CLA Coupe. Let’s set aside the obvious safety question of video meetings while driving for a second. Here’s the thing that makes me scratch my head... Mercedes is a luxury performance...
Model Failures Stem From Data, Not Algorithm Power
1/ You think your ML model fails because it’s “not powerful enough”? No. It’s your data. Garbage in, garbage out. Here’s what most AI scientists miss when using public RNA-seq or single-cell data 👇 https://t.co/dVezFRLHkW
Search Shifts From Keywords to AI‑Driven Entity Understanding
Suddenly, thousands of people in the search marketing space suddenly talk like this across all social media posts and comments: “This really encapsulates the ontological reorientation we’re seeing in search and information retrieval. We’re moving from deterministic keyword frameworks toward probabilistic,...
Acquiring Existing Startups Cuts Risk, Boosts Execution
One of the reasons people are successful acquiring startups is simple: they’re not starting from zero. Zach Simmons is a good example of this. Instead of building something new from scratch, he acquired Appraiva (@appraiva) on @acquiredotcom which already had real...
AI Sales Reps Must Match Pluribus' All‑Knowing Expertise
"AI SDRs and AEs should be as good as Pluribus. They should know everything. Every bit of detail about your product, customers, and more." with @kylecnorton CRO @owner https://t.co/bxYBgTmGEJ
Open‑weight AI + Obsidian + Crypto Enables Personal Private Programmable Stack
PERSONAL PRIVATE PROGRAMMABLE I’ve been thinking more about the intersection of Claude Code and Obsidian. There is an upcoming tech stack here that I’m calling personal private programmable. Here’s a sketch of the idea. First, if you squint ahead a few months, we...
Prompt Injection Threat Turns AI Agents Against Employers
When AI Agents Turn Against You: The Prompt Injection Threat Every Business Leader Must Understand As organizations deploy #AIagents to handle everything from customer service to financial decisions, a critical #security #vulnerability threatens to turn these digital workers against their...
Clawdbot's Power Matches Any Computer—Limited only by Imagination
Asking what you can *actually* do with Clawdbot / @moltbot is the same as asking what you can actually do with a computer. It’s capped by whatever you think of asking it. It’s controlling the whole computer.
Software and AI Profit From Margins, Not Price
Jeff Bezos said, "Your margin is my opportunity." Does this apply in software & AI? Yes—but not on price. 🧵
Founders Reveal Real AI Metrics, Pricing, GTM Strategies
At SaaStr AI 2026, no vanity metrics Founders are sharing real numbers: What AI features actually convert. What pricing actually works. What GTM motions actually scale. May 12-14. SF Bay Bring your Granola. Or even a notebook. https://t.co/BCPo2TiqOt
OpenAI Turns Into Venture Hub, Monetizing Research Tools
>be openai >announce plans to take a cut of openai-aided discoveries >release prism, free tool for research teams >rate limit prism >announce prism+, more free credits, latest models >just promise small equity if you build co >openai becomes next yc >sama used to run yc
Built‑in Agent Swarm Streamlines Multi‑Model AI Review
The agent swarm in K2.5 is a good idea. Anyone using AI seriously has pasted entire outputs into a second…third…fourth model for review and clean up. Wonder how well it works built in.
Disconnected Systems Cause Bad CX; AI Agents Can Fix
Behind almost every bad customer experience with a big company or government is tech systems that aren't connected to each other. This is one of the biggest opportunities for AI agents to improve CX.
Memorable AI Names Drive Early Virality and Adoption
some thoughts on naming AI products naming is important, and has impact, especially on early virality. babyagi was originally called "task-driven autonomous agent" and don't think it would have performed as well autogpt had a more descriptive name, and had a lot...
Continual Learning Transforms Quickly—Imagine a Year's Impact
This is what continual learning brings us: a body and a voice in a night or two. What is it going to do in a year?
Free AI-Powered Tech News Beta Launches on Pulse
Pulse is being covered on the front page of Pulse right now. How meta. 👀 Our beta launched today. Up to 50 registered users so far. Grab a free beta account and follow some topics, publishers, creators, companies, and investors...
AI Auto‑Responses: A Sales Strategy Mistake
I can't think of a worse sales strategy than auto-responding to social media posts with AI-generated comments that pitch your company. Bad look.
Interdisciplinary Communities Unlock Practical Wisdom in Responsible AI
Once in a while I write a quick post of appreciation for someone who has inspired me or my work. Today I wanted to highlight Jacob Metcalf who I went to back in 2016 as he (and his org, Ethical...
