IN A CONNECTED WORLD
Multiple perspectives. One place.
Your curiosity, your daily feed.
For you enables ranking by explicit follows and saved stories; marked-as-read history is used only with reading memory enabled. Last 48 hours first, then the last 7 days. Latest turns personalization off.
On Marketing Against the Grain, Amanda Natividad unpacks SparkToro’s 2026 zero-click study: only 32% of Google searches now earn a click, 68% end with nothing or another query, and just 232 of every 1,000 US searches still reach the open web. Zero-click rose from 45% in 2016 to 68% today, with the steepest acceleration in the last two years.
Minority Mindset’s 29-minute video maps why the Fed could hike in 2026 for three reasons and why it could be forced to cut for three pressures, then pairs each path with five asset groups through historical logic — short Treasuries, floating senior loans, energy-commodities, banks and dividend stocks for hikes; debasement hedges, housing, small and growth stocks, the broad market and speculative assets for cuts.
Released Sep 21 2026, xAI's Grok 4.7 keeps Grok 4.6's $2/$6 pricing under 200k tokens while lifting long-horizon coding and agent scores; a field test across 20 builds shows a cheap, fast model that still needs caution on detail.
The video debates whether to buy Micron before its Q4 print after a 300%+ run since fiscal Q1 2020: memory prices up 3-5x in a year, $100 billion of long-term deals through 2030 and fresh capacity from 2027 force the question of whether the classic memory cycle is breaking even if prices stall.
The Stack puts the 128GB Bosgame M5 (AMD Ryzen AI Max+ 395, Strix Halo) under the microscope: a pool five times larger than a 24GB graphics card lets huge models fit, but 256 GB/s of bandwidth dictates per-token speed. Mixture-of-Experts architectures stream far faster than dense models, long prompts hit a compute ceiling, and a price jump from $1,699 to $2,999 rewrites the value equation.
Shane Hummus pairs Bankrate's $885 average versus $200 median with the BLS five-year survival check: while crowd-hungry stores, trading and apps sink, repair, personal care and pet care hover near 60% survival. Ten gigs pass one filter — no inventory, no ad spend, a customer already asking, worst-case loss a $30 brush.
The video treats Palantir's plunge from $50 to $6 in 2021 and its later free-cash-flow surge from 2023 to 2025 as a playbook, then turns the same lens on GitLab, Samsara and monday.com, blending bull-bear debate, an eight-pillar check and a 10-year valuation tool into a price-disciplined comparison.
On a stage where Harvard clinicians meet Sadhguru, one question dominates: will AI soon substitute the psychiatrist, the psychologist and the spiritual guide? Answers oscillate between the promise of scale and the human limits recalled by touch and by soil, with a story of collapse and repair that stretches from 1920 to 2045.
Quantified Strategies tested a deliberately simple 7-day ROC mean-reversion rule on SPY from 1993 to 2026: buy at the next open when 7-day ROC closes below -3% and sell at the next open when it closes above +3%. Over 33 years the rule made 146 trades with a 71.2% win rate and a 2.40 profit factor, turning $1 into $9.15 before costs and dividends, yet a 29% drawdown and in-sample design temper live-trading claims.
The S&P 500 jumped 1.49% and the Nasdaq 2.26% to new highs, yet 60% of stocks still sit 20% below their peaks. Bora Özkent unpacks why the rally stayed narrow, why Meta's Muse agent suddenly lit up CPU names, and why forward earnings are outrunning prices.
Tech YES City connects $5,000-plus GPUs to Fed policy and a GPU-collateral debt boom: M2 is rising while QT is over, hyperscalers pour over $700 billion into silicon, and gamers foot the bill.
Open-sourced by Alibaba on 20 September, Qwen Image 2.1 packs text-to-image generation, native RGBA transparency and natural-language editing into one 7B-parameter visual checkpoint. The walkthrough shows how it runs inside ComfyUI on 8 GB and even 4 GB VRAM, hits 29 seconds per image on an RTX 5000 Ada and adds early LoRA and GGUF options — yet its research-only license remains the key blocker for production.
Raffaello Ciotta breaks down a 7-step macro model learned from an ex-JP Morgan mentor that splits direction (80% off-chart) from timing (20% on-chart). Using bank research, PMI and carry, central-bank narrative, COT and a weighting engine, he shows how the weekly GBP/NZD bias for ~180 pips was built.
Quietly shipped on September 21 at the **same pricing and velocity**, xAI's **Grok 4.7** is stress-tested in Matthew Berman's **2,819-word** deep dive against Elon Musk's prediction of parity with Opus 5; the jump from **33% to 46.3% on CursorBench 4.0**, **71% on DeepSWE**, **38% on Terminal-Bench 4.0** and **19.6% on Legal** explains why the model lands **fifth** behind Anthropic's Fable 5.1 and Astra, yet disrupts the math on **cost per completed task** at **$2 input / $6 output**.
Xiaomi's MiMo-V2.6 quietly claims the open-weights crown: Pro scores 46 on Artificial Analysis, ahead of Kimi K3 and Qwen, while Flash stays free on OpenRouter, UltraSpeed delivers up to 20x faster output at the same quality, and a 1-million-token window holds without a price hike.
On tastylive's Signal Versus Noise, unusual options flow defined the week: $17 million poured into NVIDIA and AMD calls on Friday — $12.8 million across three NVIDIA trades and a single $4.2 million AMD block. Capital chased leveraged stock replacement in the more reasonably valued leader rather than the parabolic runner.
Citi argues continual learning will lift HBM, server DDR5 and enterprise SSD demand together from 2027, leaving DRAM -8.7% and NAND -6.1% short; Stifel and TD Cowen sketch a more measured profit path, while a CXMT surprise and Monday's CPU rally show two sides of the same coin.
Silent from January to mid-July, then 181 commits in a single day at the end of August and 455 across 33 repos in September. I unpack how a two-month sprint condensed Web 4.0's AI–quantum–blockchain triad — from physical AI and voice intelligence to quantum-safe signatures — into one GitHub profile.
On Blackstone's Inside Blackstone, Amit Dixit, Winfield Sickles and Goldman Sachs President John Waldron open the same file: physical AI and robots scaling in Asia, a unanimous 25bp Fed hike and a broadening consumer recovery, then Goldman's 75/25 model, liquidity discipline in private credit and NVIDIA's $500 billion compute-financing push. Waldron puts a hiring thesis at the center: curiosity and reading still win.
Jensen Huang's pledge to sell twice as many chips next year, Meta's Muse sparking a CPU rally, and an 80% HBM4 yield plus 350% memory export surge share one tight picture.
TypeSafe's first System 1 model Jev comes from ChatGPT co-creator Diogo Almeida's team. With sub-second responses, 24x cheaper than Haiku and 230x cheaper than Fable 5.1, it brings speed and economy to Claude Code.
Pat Simmons' **4,474-word** one-shot test puts xAI's claim for **Grok 4.7** — a real rival to Fable 5.1 and Astra at a far lower price — to work across three heavy builds: an **Awwwards pixel clone, a scroll-driven 3D keyboard page and a Rainbow Road Mario Kart**. With **46% vs 51.6% on CursorBench**, **71% vs 70% / 72.7% on DeepSuite**, **$2 / $6 vs $10 / $50** per million tokens and **$7 vs $123 / $97** per award build, the sheet shows why cheap does not carry heavy code.
Micron is set to report its biggest quarter on September 30, guiding to about $50 billion in sales and roughly $31 per share — nearly ten times year-ago levels — yet the street is already a notch higher and options imply an 11% move.
Former CFTC Chair Chris Giancarlo told Paul Barron about his Oct 27 book and the collapsed Clarity bill in Washington. Banks' reflex to block innovation, the issuer edge of tokenized equities, DTCC's Canton move and the crisis of KYC surveillance all came together in one conversation.
Among the noise, meaningful connections.
