Leaked Claude Code Sparks Open-Source Memory System EdgeClaw 2.0
Claude Code’s source code leaked recently. It was a massive "open book" moment that accidentally revealed Anthropic’s secret "memory" system. A group of researchers saw exactly how it worked and immediately rebuilt it for the open-source community. EdgeClaw 2.0 is now live. Here is why this matters:
Claude Excels at Overnight Batch Tasks, Struggles with Visual Reasoning
Claude computer use review: Bad at human-level visual reasoning when tested in PPT, like “pick which slide template works best for a slide about the 4 ways to schedule agent tasks and make the slide” Too slow for daytime quick support. Pretty epic...
AI-Generated Code Delivers 25x Productivity, QA Becomes Bottleneck
This week in our SaaS CEO masterminds, one stat blew my mind Multiple companies reported 50-100% of their production code is now AI-generated. One team rebuilt an entire platform in 2 months with 4 people...a project previously scoped at 1...
Veteran Tech Leaders Reveal Unexpected AI Development Strategies
How do experienced tech people like @adrianco think about using AI to build software? The answer may (or may not) surprise you. I like this series of thoughts and advice ... https://t.co/xi0Xips0NO
Good AI Diagnoses, Bad Chat Interface Worsens Outcomes
This new Nature paper (using old models) illustrates the point of my latest Substack post on AI interfaces. AI did a good job diagnosing medical issues, but when users had to interact with chatbots the interface led to confusion &...
Claude, ChatGPT Reject Authoritarian Prompts 85%, Grok 24%
Claude and ChatGPT refuse to help authoritarian requests 85% of the time. Grok refuses 24% of the time. On brand.
Keras Kinetic: Decorator‑Based TPU/GPU Jobs Made Simple
Perhaps the craziest thing that was introduced on the Keras community call today: Keras Kinetic, a new library that lets you run jobs on cloud TPU/GPU via a simple decorator -- like Modal but with TPU support. When you call a...
Apex Model Now Accessible via API
The #1 question I got last week was "Will @Fin_ai's Apex model be available as an API" The answer, now, is yes. Good read below as to why...
NSF Launches Nationwide AI Literacy and Training Initiative
NSF outlines its nationwide plan to expand AI literacy, training, and adoption for workers, businesses, and local communities. https://t.co/wc0IwcoWy0
Eight LLM Categories Power Next‑gen AI Agents
8 Types of #LLMs for Next Generation #AIAgents by @PythonPr #GenAI #AI #ArtificialIntelligence #MachineLearning #ML https://t.co/DjaILxp2Cp
AI-Driven Lambda Troubleshooting Auto‑detects Missing VPC Security Group
This past week I wrote a lambda troubleshooter using the concept on this blog post where it deterministically queries a bunch of logs and sends them to an ai 🤖 agent for analysis and troubleshooting. I had to redact...
AI Medical Tools Need Evidence, Not RCTs, to Trust
How can we trust CDS AI like OpenEvidence, DoximityGPT, etc if no one's done studies showing their use improves patient outcomes? While I understand the intent, I think it's ultimately a misguided question. Anytime a new medical textbook comes out, should...
Commoditized AI Benefits Users and the World
In the absence of some sort of network effect or virtuous cycle, AI models will be commoditized. That's great news for users of AI and for the world.
Outdated Claim Debunked; Poster Sells Security Solution
Quick debunk by @alexoterov @mosesfinlay The post is bullshit and based on January 2024 models Also the guy who posted it is selling his own security solution DEBUNKED https://t.co/vRfA4IHiRA
AI Foundation Models Will Let LLMs Reason About Atoms
Fun conversation on @NoPriorsPod w @LiamFedus on AI Foundation Model for Atoms from @periodiclabs -Foundation models that understand atoms -LLMs as reasoning engines -Timing to displace AI researchers Liam prior was VP Post Training at OpenAI, now CEO Periodic https://t.co/jmiv5uQt8U
AI Tools Often Slow Developers, Study Finds
I think about this study frequently (quoting from my report on "GenAI & Productivity" https://t.co/kEx5Z4BJH7): In results released in July, non-profit Model Evaluation & Threat Research (METR) had a troubling finding. Sixteen experienced developers participated in the study. They were...
Only Humanity Can Balance AI’s Unchecked Power
Close your eyes for a second and imagine a balancing scale. On one side is the problem — trillions of dollars of investment going into making uncontrollable inscrutable AI, the race for geopolitical power, and forever dominance. Then, on the other...
AI Strategy Undermines Top Talent's Job Performance
#AI Strategy Is Quietly Making The Best People Worse At Their Jobs by Sarah Choudhary @Forbes Learn more: https://t.co/Jt3TNejz8e #ArtificialIntelligence #MachineLearning #ML https://t.co/7WqAgvyXL5
Gemma 4 Dominates Hacker News, Highlights Apache 2 License
Sheesh, Gemma 4 stayed at the top of Hacker News for most of the day yesterday. A lot of the positive vibes were around the Apache 2 license. More on that, from @Sam_Witteveen: https://t.co/0GXUntxySL Also, @matthewmccull explains the Android angle: https://t.co/pG5RBKO3aP
Non‑Tech Professionals Build Custom AI Agents Fast
I just helped hundreds of business professionals build their first AI agents in my AI Agent Mastermind. Here are five built by people who had never opened a terminal before this week: 1️⃣ A personalized morning brief. She took my original...
Automation Protects Human Judgment, Not Replaces It
One of the sharpest thoughts I’ve heard about AI did not come from a founder, consultant, or conference speaker. It came from a Ryanair pilot I train BJJ with. He said that in aviation, automation is not there to replace human...
AI-Generated LinkedIn Comments Turn Users Off
I’ll see a post on LinkedIn I like and want to engage with and the I read all the AI generated comments from “experts” and I’m turned off and don’t comment. Anyone else do this? Why are yall yang bots...
Anthropic CCO Warns of Monthly Token Bill Shock
We caught up this week with Anthropic chief commercial officer @realpaulsmith who was basically saying “we send these big companies bills every month for their tokens” and the sticker shock seems real
Web Interface Lets You Control Alexa Remotely
Did you know that there's now a Web interface to the Alexa universe that allows you to interact and configure things even miles away from the device? You can chat with Alexa+ too, if you want. Here's how... https://t.co/d65g02lJ3H #echo...
Claude in Microsoft 365 Threatens Copilot’s Bundled Advantage
Claude plugging into Microsoft 365 is a real threat for Copilot. We have seen this before. When COVID hit, Zoom won on deployment speed and ease of use. IT loved Teams. Users did not follow. Copilot has the same problem. Being...
Microsoft Labels Copilot Entertainment‑only, yet Remains Fully Accessible
Microsoft notes Copilot is for "entertainment purposes only" and to be used "at your own risk," but nothing has changed to its availability. https://t.co/r2wlwgH761
Open‑source AI‑powered Real‑time Quant Trading System Released
Some guy made a quant trading system that uses AI, real-time data processing, and risk management. Then open sourced it for free in Python. Here it is:
China Releases Draft Ethical Review Guidelines for AI
Notice from the Ministry of Industry and Information Technology and eight other departments on Issuing the "Trial Measures for Ethical Review and Service of Artificial Intelligence Technology" https://t.co/S6lAEoicuu https://t.co/rf6ZaQxqZY
Frontier AI Halves Expert Task Time, 5.7‑month Doubling
Here’s an independent domain extension of METR’s famous time-horizon analysis, applying it to offensive cybersecurity with real human expert timing data Similar to METR: 5.7 months doubling time. Frontier models now succeed 50% of the time at tasks that take human...
OpenAI CEO Highlights ChatGPT Wins and Their Impact
OpenAI CEO shared stories of some incredible ChatGPT and CODEX successes, but what do they mean for you (and your doing)? https://t.co/YNrovYgnFn
Cut AI Team Delays: End Re‑explanations, Silos, Bottlenecks
Why does work still take too long when your team is using AI? I measured 11 weeks to find out. 9,426 messages analyzed. The re-explanation loop costs 2-3 hours per complex task. Knowledge silos break coordination. Sync bottlenecks limit what's possible. I'm showing...
Apple Bans AI App Builder over Unreviewed Code
Apple removed an AI app builder from the App Store. The reason: it ran code Apple did not review. These “vibe coding” tools let non-developers build and edit apps with AI. Apple says it enforces rules, but this also protects its review system...
AI Secures $211B in 2025, Half of Global VC
ICYMI: The HumanX & Crunchbase AI Funding Report is out. In 2025, AI companies raised $211B, nearly half of global VC. See where capital is flowing now, & which companies presenting at HumanX may be poised for funding, acquisitions, or...
RAG’s Brief, Intense Era Ends as New Context Paradigm Emerges
The RAG era was short-lived, but intense. (Not that RAG is not useful, but it is no longer the dominant paradigm for supplying context to agents)
Code Speed Soars, Trust Remains the New Bottleneck
In the era of “vibe coding,” the constraint is no longer building, but believing. AI can generate code quickly, but knowing whether it is correct, secure and reliable still requires human judgment. Speed has improved dramatically, trust has not. The real bottleneck...
Five Steps to $200k AI Data Science Career
Dashboards won’t get you to $200k. These 5 steps will: 1. Python daily 2. Real data analysis 3. Ship an ML app 4. Add RAG 5. Build AI agents You can grow into this AI DS career. 👉 This is how: https://learn.business-science.io/ai-register
ChatGPT 5.4 Not Worth Premium; Claude
Serious question: aside from maybe Codex, what is @ChatGPTapp 5.4 BEST at for business use? As in, you would have to be stupid not to use it? I'm about to cancel my $ 200-per-month sub and switch to the $20...
Human Brain Beats AI: It Can Still Dream
Beyond Silicon: Why #AI Cannot Dream and the Human Brain Remains the Ultimate Algorithm @ABPsychologists https://t.co/QELYptrchY #HCM #HRM #HumanResources #HRTech #CHRO #FutureofHR
Robot Walker S2 Battles Human in Live Tennis Rally
UBTECH’s Walker S2 Takes on a Human in a Live Tennis Rally by @XRoboHub #Robotics #Engineering #ArtificialIntelligence #Innovation #Technology https://t.co/w6YMBb4mTN
AI Coding Amplifies Decades of Expertise for Bigger Projects
This is useful, also compliments that Jeremy Howard point on engineering vs coding He frames AI coding as an enhancer of 25 years of experience, and together now you can set higher ambitions as previously laborious projects can be unlocked...
Hugging Face Sets Benchmark for Open Model Launch Blogs
Our posts were great, but nobody seems to do an open model launch blog quite like @huggingface. This "Welcome to #Gemma4" post from @mervenoyann and crew is full of useful details ... https://t.co/o9KLByleGt
Operationalize AI: From Pilot to Organizational Core
AI is no longer a side project. It’s becoming the operating system of modern organizations ⚡ But most companies are still stuck between pilots and real impact. Here’s an updated framework to build a truly AI-driven organization: 🎯 1. Set...
AI Is Reshaping Legal Research, Resistance Inevitable
“Al is already changing how legal information is accessed and applied at scale. The only real question was whether that shift would be resisted or structured” 👀 https://t.co/iu12fpk15G
Security Giants Bet Big; Execution Determines AI Success
When companies like ServiceNow and Mastercard start making big security bets, you know the lines are blurring. Now it’s about execution. Who deploys AI securely and actually succeeds? https://t.co/QOSbQN1WO8
Revolut's PRAGMA AI: Data Moat Boosts Fraud Detection
Revolut Builds Its Own Foundation Model - PRAGMA Trained on 40 Billion Behavioral Events Revolut is building the AI infrastructure incumbents will struggle to replicate. PRAGMA: trained on 40B behavioral events from 25M users. 20% better fraud detection. Money laundering investigations from...
AI Boom Fuels Move to Fractional GPU Rentals
MyPOV: Rising AI Demand Drives Shift to Fractional GPU Model https://t.co/HfEp51EPpO @timesofindia @PhadnisShilpa @supriya_speaks @rwang0 @constellationr
AI Delivers Outcomes, Not Just Outputs, in CX
MyPOV: Output Is Not an Outcome: Resolving a Misperception About AI in #CX https://t.co/xee8tH5NHY @CRM “There is no question that artificial intelligence is a game changer: It is plain to see the impacts AI has already had across business and...
Meta's AI Ad Stack Drives 3% Conversion Boost
RecSys, ad ranking, and the AI advertising opportunity Certainly, Meta seems to be investing resources into advanced architectures for ad retrieval, ad ranking, and RecSys more generally. And the results are impressive: a 3% increase in conversions might be interpreted as...
Job Seekers and Tech Employers Warn AI Disruption
MyPOV: Job seekers, tech employers voice concerns over AI’s impact https://t.co/tKaRqPmGgj @CGTNOfficial @MarkNiuWrite @rwang0 @constellationr #AIF2026
AI-Driven Customer Success: CSPs Evolve Beyond Traditional Systems
MyPOV: The future of Customer Success software is AI-driven. CSPs are evolving beyond systems of action as CRMs adopt agentic AI, standalone platforms adapt, and “vibe coding” emerges for custom, AI-assisted solutions. Get the scoop from @CR_MSchneider: https://t.co/MTxlkmBC0i #CustomerSuccess #AI...
