# 4x (blogs) — RSS Amplifier Recent posts from the 2 feeds in the RSS Amplifier directory that cover 4x. Page: Feed: --- ## [I forced a Mixture of Experts to specialize by topic. It got worse.](https://www.phenx.io/post/forced-mixture-of-experts-specialize-by-topic) _2026-07-28 · saurabhsarkar · Phenx Inc_ The folk model says a Mixture of Experts holds a code expert and a medicine expert. Trained from scratch on two consumer GPUs, the routing tells you more about whether a word is a comma than about what the document is about, and it froze in the first 2% of training. Three attempts to force topic routing all came out slightly worse. ## [What does a Mixture-of-Experts router actually read?](https://www.phenx.io/post/what-does-a-mixture-of-experts-router-actually-read) _2026-07-22 · saurabhsarkar · Phenx Inc_ The folk model says a Mixture of Experts has topic specialists: a code expert, a geography expert. I aimed a Jacobian lens at the router on OLMoE and Qwen-MoE and read what actually flows into each expert. It sorts brackets, newlines, and verb tense, not topics. ## [Mixture of Experts: the sparse model that lights up 93% of itself](https://www.phenx.io/post/mixture-of-experts-the-sparse-model-that-lights-up-93-percent-of-itself) _2026-07-20 · saurabhsarkar · Phenx Inc_ A Mixture-of-Experts model is supposed to spend a fraction of itself on each token. I put CUDA events around every MoE layer of OLMoE-1B-7B on an RTX 4090 and recorded which of its 1,024 expert slots actually fire. At a batch of eight, 628 of them fire every single step -- and at batch 32 it is 93%. The sparsity is real, it is just not where people think it is. ## [J-Scope: before a model answers, it thinks in concepts](https://www.phenx.io/post/j-scope-before-a-model-answers-it-thinks-in-concepts) _2026-07-15 · saurabhsarkar · Phenx Inc_ The prompt never says Italy, but twelve layers into Qwen3.5-4B the model is already thinking it. J-Scope lets you watch that happen, inject France and get Paris, or drag one vector and watch the answer walk from euro to yen. A 72-second tour here; the live tool is one click away. ## [How to mine iMessages to reactivate dormant relationships and build a personal AI voice](https://moxiegtm.substack.com/p/how-to-mine-imessages-to-reactivate) _2026-07-14 · Richard F. Purcell · MoxieGTM_ Your texts are a unique dataset and its sitting in your pocket waiting to be mined. ## [I Fed a Model Facts Through Its J-Space Instead of Its Prompt](https://www.phenx.io/post/i-fed-a-model-facts-through-its-j-space-instead-of-its-prompt) _2026-07-08 · saurabhsarkar · Phenx Inc_ I wrote facts directly into a 4B model's J-space and KV cache on an RTX 4090. Three cache tokens matched full text-RAG, survived 48 distractors, and showed where retrieved facts actually live. ## [Can you tell a good AI agent plan from a bad one before you run it? I spent 6 million tokens finding out.](https://www.phenx.io/post/can-you-tell-a-good-ai-agent-plan-from-a-bad-one-before-you-run-it-i-spent-6-million-tokens-finding) _2026-07-02 · saurabhsarkar · Phenx Inc_ Can a cheap reward model tell a good AI agent plan from a bad one before you run it? I reproduced Orch-RM on a consumer RTX 4090 and RTX 3090: the verifier barely beats a coin flip on the signal that matters, yet still beats majority vote. Cheap to run, expensive to teach. ## [I Gave My Coding Agent Karpathy's Discipline Rules. It Got Too Careful to Fix the Bug.](https://www.phenx.io/post/i-gave-my-coding-agent-karpathy-s-discipline-rules-it-got-too-careful-to-fix-the-bug) _2026-06-28 · saurabhsarkar · Phenx Inc_ I tested the viral Karpathy coding-discipline rules on 100 real SWE-bench bugs, three times. They didn't help: the agent fixed fewer of them, trading fixes for caution. ## [How to scale referrals in half the time and 10% of the cost](https://moxiegtm.substack.com/p/how-to-scale-referrals-in-half-the) _2026-06-26 · Richard F. Purcell · MoxieGTM_ Who gets the credit when an SDR books a meeting off a list Claude built: the bot or the human? ## [I bolted MiniMax's MSA sparse attention onto a 3B model on a single 4090](https://www.phenx.io/post/i-bolted-minimax-s-msa-sparse-attention-onto-a-3b-model-on-a-single-4090) _2026-06-25 · saurabhsarkar · Phenx Inc_ MSA sparse attention, retrofitted onto a 3B model on a single RTX 4090 with no training: the 28.4x compute cut holds, quality survives, plus caveats the paper skips. ## [Trade Smarter With IBKR (Sponsored)](https://crawlproof.com/a/5QMb1IX3xMAQ) _2026-06-25 · **Sponsored**_ Join Interactive Brokers through a referral and start investing with low-cost global access. ## [I ran DFlash on a MacBook. The 4x is a data-center number.](https://www.phenx.io/post/i-ran-dflash-on-a-macbook-the-4x-is-a-data-center-number) _2026-06-23 · saurabhsarkar · Phenx Inc_ I tested DFlash speculative decoding on an M5 Max and RTX 4090. The 4x headline is real, but workload and concurrency decide whether you get a speedup or a slowdown. ## [Why your adaptive controller might be causing the defects it’s trying to prevent](https://www.phenx.io/post/why-your-adaptive-controller-might-be-causing-the-defects-it-s-trying-to-prevent) _2026-06-11 · saurabhsarkar · Phenx Inc_ A field note on over-cautious process control, and the margin hiding in your own sensor data. In one engagement: the force the controller watches stays low across speeds, while sidewall wear spikes at the crawl. “Slower is safer” is the most expensive assumption on your shop floor. Most adaptive controllers are built on it. When cutting force climbs, the controller slows the tool; when force… ## [The Geometry of Intelligence: What Lagrange Can Teach Modern AI](https://www.phenx.io/post/the-geometry-of-intelligence-what-lagrange-can-teach-modern-ai) _2026-05-01 · saurabhsarkar · Phenx Inc_ In 1772, Joseph-Louis Lagrange solved a piece of the three-body problem that had stumped everyone before him. He didn't solve it by tracking forces. He solved it by finding the five points in the gravitational field of two large bodies where a third, smaller body can sit stably - held in place not by any single pull, but by the balance of all of them. These are the Lagrange points. They are where… ## [How to train an AI agent the same way you'd train a Virtual Assistant](https://moxiegtm.substack.com/p/how-to-train-an-ai-agent-the-same) _2026-04-13 · Richard F. Purcell · MoxieGTM_ If you can dodge a wrench, you can dodge a ball. If you can train a VA, you can train an AI agent. A 7-part how-to guide. ## [How to cancel your Clay subscription](https://moxiegtm.substack.com/p/how-to-cancel-your-clay-subscription) _2026-03-19 · Richard F. Purcell · MoxieGTM_ 15 minutes could save you 90% or more on your data enrichment ## [How to 2X followers and generate 4X ROI from your LinkedIn brand](https://moxiegtm.substack.com/p/how-to-2x-followers-and-generate) _2026-03-05 · Richard F. Purcell · MoxieGTM_ From 4,500 to 10,000 followers: Updating the ROI on an up-and-coming LinkedIn Brand ## [How to use Claude Cowork to 2X performance and reduce 90% of admin time and tech investments](https://moxiegtm.substack.com/p/how-to-use-claude-cowork-to-2x-performance) _2026-02-23 · Richard F. Purcell · MoxieGTM_ Claude Cowork is the new frontier for AI agents just like ChatGPT in 2022 was the new frontier for LLMs ## [How to 4X revenue by identifying the right offer](https://moxiegtm.substack.com/p/how-to-4x-revenue-by-identifying) _2026-02-11 · Richard F. Purcell · MoxieGTM_ The wrong offer can be the difference between a 20% or a 30% close rate ## [How to use data (not vibes) to validate new markets](https://moxiegtm.substack.com/p/how-to-use-data-not-vibes-to-validate) _2026-02-03 · Richard F. Purcell · MoxieGTM_ Why choosing markets based on pipeline or cultural will kill your business ## [How to go from $30K/month to $0/month in 4 months and then pivot to close 3 deals in 45 days](https://moxiegtm.substack.com/p/how-to-go-from-30kmonth-to-0month) _2026-01-30 · Richard F. Purcell · MoxieGTM_ A successful pivot still couldn't save a business in an unwinnable market ## [Real SIMs for AI Agents (Sponsored)](https://crawlproof.com/a/U7zJylwvOIMd) _2026-01-30 · **Sponsored**_ Pass 2FA and verifications with real carrier numbers via our API. ## [How to use AI to sound more human & generate 10%+ reply rates](https://moxiegtm.substack.com/p/how-to-use-ai-to-sound-more-human) _2026-01-20 · Richard F. Purcell · MoxieGTM_ Use AI to accelerate creativity and identify who is worth a human touch. ## [The Real State of Generative AI in the Enterprise in 2025](https://www.phenx.io/post/the-real-state-of-generative-ai-in-the-enterprise) _2025-12-12 · saurabhsarkar · Phenx Inc_ Why most enterprise GenAI initiatives stall after pilots and what actually scales The Investor Narrative vs the Operator Reality From the outside, enterprise generative AI looks like a straight line. Budgets are approved. Pilots launch quickly. Demos look impressive. Board decks show steady progress. Recent industry benchmarks show broad enterprise investment and deployment of generative AI, with… ## [How AI Is Transforming the Financial Industry](https://www.phenx.io/post/how-ai-is-transforming-the-financial-industry) _2025-11-11 · saurabhsarkar · Phenx Inc_ Artificial intelligence is no longer a futuristic concept reserved for sci-fi movies. It’s here, and it’s shaking up the financial world in ways we never imagined. From speeding up transactions to detecting fraud before it happens, AI is transforming finance at a breakneck pace. But how exactly is this happening? And what does it mean for the future of money management? Exploring AI Applications… ## [When Computation Surprises Itself: How Emergent Uncomputability Might Explain AI, Consciousness, and the Universe](https://www.phenx.io/post/when-computation-surprises-itself) _2025-11-05 · saurabhsarkar · Phenx Inc_ How Emergent Uncomputability Could Explain Both the Universe and Artificial Intelligence The idea that we might be living inside a simulation refuses to die. It’s one of those questions that blends philosophy, physics, and computer science so elegantly that it keeps resurfacing in every generation: If reality follows rules, could those rules be running somewhere? When a team of physicists from the… ## [Revolutionize Operations with Custom AI Solutions](https://www.phenx.io/post/revolutionize-operations-with-custom-ai-solutions) _2025-11-04 · saurabhsarkar · Phenx Inc_ Imagine a world where your business operations run smoother, faster, and smarter. Sounds like a dream, right? But with the rise of artificial intelligence, this dream is becoming a reality. The key? Tailored AI solutions designed specifically for your unique challenges and goals. Let’s dive into how these customized technologies can transform your operations and give you a competitive edge. Why… ## [Advanced Fraud Detection Solutions for Your Business](https://www.phenx.io/post/advanced-fraud-detection-solutions-for-your-business) _2025-10-27 · saurabhsarkar · Phenx Inc_ Fraud is a sneaky adversary. It evolves, adapts, and strikes when you least expect it. If you think your business is too small or too niche to be targeted, think again. Fraudsters don’t discriminate. They look for vulnerabilities, and if you’re not prepared, you’re an open door. So, how do you stay ahead? How do you protect your assets, your reputation, and your customers? The answer lies in… ## [When a Few Pixels Redrawn Change Everything: The Battle Over Human Authorship in the Age of AI](https://www.phenx.io/post/intellectual-property-and-ai-generated-content-a-first-principles-view) _2025-10-18 · saurabhsarkar · Phenx Inc_ Understanding Intellectual Property in the Modern Age At its core, intellectual property (IP) is not merely about ownership of ideas. It revolves around incentive design . Societies discovered long ago that creativity and invention flourish when individuals can reap benefits from their mental labor. However, pure ideas are non-rivalrous: once shared, they can be copied infinitely. Therefore, IP… ## [Animals vs Ghosts by Karpathy: A Realistic Path for Enterprise AI](https://www.phenx.io/post/animals-vs-ghosts-a-realistic-path-for-enterprise-ai) _2025-10-07 · saurabhsarkar · Phenx Inc_ Animal vs Ghost Most AI systems today are ghosts: trained on human text, fine-tuned for alignment, and limited to observing and... ## [Finance x AI - Key Trends and Breakthroughs in 2025](https://www.phenx.io/post/finance-x-ai-key-trends-and-breakthroughs-in-2025) _2025-09-30 · saurabhsarkar · Phenx Inc_ Here’s what’s changed or accelerated in recent months, the “hot takes” you can’t ignore. 1. AI Agents + Multimodal Intelligence Are... ## [International money transfers (Sponsored)](https://crawlproof.com/a/9EF8BdXIEdar) _2025-09-30 · **Sponsored**_ Send money to other countries and manage multi‑currency balances online. ## [Tailoring AI to Meet Business Needs](https://www.phenx.io/post/tailoring-ai-to-meet-business-needs) _2025-09-22 · saurabhsarkar · Phenx Inc_ Artificial intelligence (AI) is transforming the way businesses operate across industries. However, the true power of AI lies not in...