Souk's Notes - Unofficial Summaries

2026-02-21 · open the original ↗

• Agents now pick dev tools by parsing documentation, not word of mouth. YC partners argue this is reshaping dev-tool go-to-market, citing a surge in Postgres databases created over the past year as agents building apps default to Supabase, whose documentation they say is simply the best-written for an LLM to parse.
• Resend rebuilt its docs to be agent-readable, and it worked. The email-sending company's founder noticed over a year ago that ChatGPT had become a top-three inbound channel, so he restructured docs into bullet-point Q&A with embedded code snippets; partners contrast this with SendGrid's support-ticket-gated docs, which they say leave it invisible to agents.
• Documentation infrastructure is becoming a must-have, not a nice-to-have. Partners point to Netlify, built on better API documentation, as newly positioned to capture "exponentially more agents making exponentially more decisions" than developers ever did — meaning even a small documentation improvement could meaningfully move a dev-tool company's business.
• Agent Mail's agent-only inboxes took off once OpenClaw exploded. The YC company built email accounts designed for AI agents because Gmail and other providers deliberately obstruct automated sign-ups to fight spam; partners speculate this points toward a fuller "agent economy" — even agent-native phone numbers — running alongside, and possibly eventually distinct from, human commerce.

2026-02-21 · open the original ↗

• Quantum could turn approximate routing answers into optimal ones. Murugesan notes the travelling salesman problem is NP-hard: one vehicle visiting 20 stops has, by his uncertain recollection, well over a trillion orderings, and today's methods rely on approximation and shrinking the search space. The host quotes the guest's Diginomica article saying work that would take a classical platform 10 septillion years could be done in minutes with quantum.
• Real routing compounds the problem far beyond a single tour. He scales it up to the vehicle routing problem (10,000 stops, 15 vehicles), then adds the choice of road segments between stops and real-time traffic, which current approximation models mostly predict rather than ingest. Abundant compute would allow much higher precision, though he grants a point of diminishing returns on what is "good enough".
• Self-operating logistics remain far off. He expects today's technology to keep supporting assisted decisions and says "I don't think we're anywhere close" to truly self-operating systems, with compute and real-time data access the limiting factors.

2026-02-21 · open the original ↗

• Investment management is ripe for automation by algorithms. Morfit says the industry is full of inefficiencies. Asset allocation, tax optimization, hedging and life-stage planning are highly programmable for both individuals and institutions. Meanwhile, building products to meet those needs has been fragmented and bespoke.
• The ETF manager is really a data and software company. Long seen as a diversified asset manager good at ETFs, BlackRock caught his attention over the past 12 months for its Aladdin platform. Aladdin manages positions and risk and is being put on the desktops of wealth and asset managers across the industry, driving adoption of software-run model portfolios.
• Owning the AI control plane would be a commanding position. Morfit argues that in an AI world, the control plane where decision-makers deal with software agents will be critical. If that interface is Aladdin, BlackRock would be in an "incredibly advantageous position." This is his reason for the new position.

2026-02-21 · open the original ↗

• He describes Biden as visibly more physically diminished than other leaders of similar age at the July 2024 NATO summit. He frames this as a medical observation rather than a partisan one, saying he wants a leader who can speak coherently regardless of party.

2026-02-21 · open the original ↗

• A "cup and handle" chart pattern formed on Bitcoin running from April 2021 to the week of November 4, 2024, which the analyst reads as bullish if it breaks out. The cup's low is placed at $64,859 and its high at $68,976, a band he calls psychologically significant since Bitcoin is currently trading back in that range.

2026-02-20 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2026-02-20 · open the original ↗

• Claude Code's creator didn't know at first that he had something, then became consumed building it. He describes working every night and weekend for three straight months without a vacation once he sensed "this is going to be a thing," even though it couldn't yet code; he says Anthropic's internal principle — build for the model six months out, not for today's — remains his advice to founders building on LLMs.

2026-02-20 · open the original ↗

• The Supreme Court's 6-3 ruling struck down most Trump tariffs, but roughly $130 billion already collected under IEEPA won't simply flow back to consumers. Refunds depend on whether an import entry has been "liquidated" (finalized) by Customs — large importers mostly protected themselves by filing protective lawsuits to keep entries open, while smaller businesses that couldn't afford to are largely locked out of refunds.
• The fallout splits unevenly by company size and supply-chain position. Only the importer of record can claim a refund, so manufacturers, distributors, and retailers who absorbed passed-through tariff costs downstream have no direct claim against Customs, setting up a wave of contract disputes and finger-pointing lawsuits over "tariff adjustment clauses" up and down supply chains.
• The government's likely defense is prospective-only relief, but tax-refund precedent cuts against it. Officials will argue liquidated entries can't be refunded, yet the author notes retroactive refunds are typically required for invalidated taxes, and tariffs fall under Congress's taxing authority — while Trump has already signaled a pivot to reimpose 10% tariffs under statutes the Court didn't touch (China trade practices, steel/aluminum).
• The regressive real-world cost undercuts the stated rationale for the tariffs. U.S. households reportedly lost $1,000–$2,400 in purchasing power in 2025 (Yale Budget Lab, Tax Foundation), with apparel, groceries, electronics, and cars all rising; the author argues tariffs functioned less as an economic strategy than as negotiating leverage and theater, hitting lower-income households hardest since they spend more of their income on the goods affected.

2026-02-20 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2026-02-20 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2026-02-20 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2026-02-20 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2026-02-20 · open the original ↗

• The AI-jobs anxiety is rational but leaves out the evidence. The host lists its sources: AI entrepreneur Matt Shumer's viral post "Something Big Is Happening", claiming one can describe an app in plain English, walk away for 4 hours and return to a finished, tested product; Anthropic CEO Dario Amodei's prediction that AI could eliminate half of entry-level white-collar jobs within 1 to 5 years; the IMF's Kristalina Georgieva's "tsunami"; and Jamie Dimon saying JPMorgan Chase would soon need fewer employees. He offers The Economist's article "Why AI won't wipe out white-collar jobs" as counterweight.
• US white-collar jobs and pay have grown since ChatGPT launched. Per The Economist, the US added roughly 3 million white-collar jobs since late 2022 while blue-collar employment stayed flat, and it has 7% more software developers, 10% more radiologists and 21% more paralegals than 3 years ago. Real wages rose about 5% in professional and business services and 9% for office and admin workers, and the controlled white-collar premium of roughly a third is nearly triple its early-1980s level and still rising.
• Technology automates tasks within jobs, so jobs mostly get upgraded. The Economist recalls Wassily Leontief's 1982 warning, after which management, professional, sales and office employment more than doubled and real pay rose about a third. Its analysis of more than 100 occupations since late 2022 finds project managers and information-security experts up around 30%, insurance claims clerks down 13% and secretaries and admin assistants down 20%, while unnamed "other mathematical science" occupations grew around 40% with real wages up roughly a fifth; the host adds Daron Acemoglu and Pascual Restrepo's finding that about half of US employment growth from 1980 to 2007 came from occupations gaining new titles and tasks.
• Entry-level and clerical roles are exposed, and capability is compounding. Clerical and admin work fell from 18% of Americans in the 1980s to 10%, and research by Sam Manning and colleagues at the Center for the Governance of AI finds clerical workers least able to adapt. METR data show the task length frontier agents complete with 50% reliability doubling about every 7 months for 6 years, now over 6 hours of expert work, which the host extrapolates to multi-day autonomous work about a year away. Against that, Anthropic's usage data show only about 4% of occupations using AI across three-quarters or more of their tasks, which The Economist ties to "jagged intelligence".
• The host's advice: stay technical and learn where AI fails. His four recommendations are not to panic out of a technical career, to become the person who works with AI (The Economist's "cyborg workplace"), to invest in judgment, stakeholder communication, coordination and messy domain expertise, and to keep experimenting because today's models will be eclipsed in 6 months. He concedes real disruption in back-office and some entry-level work but calls the claim that all cognitive work is ending unsupported by employment, wage, historical or usage data.

2026-02-20 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2026-02-20 · open the original ↗

• Multi-step AI features leak users out of the treatment. In Shane's illustrative case, an internal AI analyst V2 is tested on 1,500 daily users. Of the 743 assigned to it, 156 never open the tool, 190 hit cached V1 results, 129 time out to a V1 fallback and others fall back on errors, leaving 167 (23%) who actually got V2. The headline +3% (p = 0.15) hides roughly +14% among those exposed, but with 167 users against a power calculation needing 760, the test has to be fixed and rerun rather than filtered after the fact.
• AI tests need far larger samples than the usual calculation suggests. Shane says a diluted effect needs disproportionately more users ("half as loud" means listening four times as hard). Non-deterministic outputs add variance, and performance that differs by segment, such as simple versus complex queries, needs a full sample per segment. Because these multiply, he says a test can need around 10 times the users expected.
• Measure correctness with code where possible, and calibrate offline evals. "SQL success rate" can mean valid syntax, executed, non-empty or actually correct. In the example, checking V2 against 200 finance-maintained reference answers finds 5% confidently wrong (off by more than 10% behind a convincing chart). An LLM judge that agrees with humans 80% of the time is a "noisy pen" that washes out real differences, so Shane's rule is to use deterministic code checks for everything else. He argues that over 5–15 experiments, offline quality gains can be calibrated to online outcomes such as task completion.
• More guardrails and pre-registered segments change the ship decision. Shane lists guardrails beyond latency and crash rate: cost per query, safety and PII violations, hallucination, confidently-wrong rate and data leakage, each needing a measurable online mechanism, with quality gains often costing latency or money. Averages hide failures, so the call may be to ship for simple queries and hold for complex ones; segments must be chosen before the test to avoid sample-size shortfalls and p-hacking.
• Run AI tests longer, because users change what they ask. Beyond novelty effects, Shane argues that users who come to trust an AI analyst move to harder questions, so a query mix of 80% simple / 20% hard may shift to 50/50 after launch. He has no fixed rule but suggests four weeks or more and a 5% holdout for a couple of months. In Q&A he describes his team's "shadow mode", running V2 silently alongside V1 on production traffic for realistic quality metrics, which he would build earlier next time; his legal-AI team also leans on causal inference and opt-in beta users because lawyers lose trust fast.

2026-02-20 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2026-02-20 · open the original ↗

• The stock market is best understood as an auction tracking where money flows and why, not a verdict on value. He gives the example of buying Apple on a bet it wins consumer AI: more buyers than sellers pushes a stock up, and the same logic applied to currency demand explains dollar strength.

2026-02-20 · open the original ↗

• The guest argues the bond market, not the stock market, is what actually disciplines the president now. He contrasts this with Trump's first term, when tariff policy set in 2016 was rolled back once the S&P 500 fell to a given level in response — a stock-market check he says is not happening this time.

2026-02-20 · open the original ↗

• The guest, finance analyst Jessica Inskip, argues the bond market now drives policy more than equities, calling the 10-year Treasury yield the president's "daddy." She says Trump's first term rolled back tariffs when the S&P 500 hit certain levels, but this term the trigger is the 10-year yield, which she ties to mortgage rates and the political need to keep the economy looking healthy in a midterm year; she notes the Fed only directly controls the shorter end of the curve.
• The guest frames stablecoins as a structural fix for fading foreign demand for US Treasuries, itself a byproduct of tariffs reducing the dollars America exports. She explains foreign holders traditionally recycled trade-surplus dollars into Treasuries, but tariffs shrink that flow; she points to the "Genius Act," which requires stablecoins to be backed one-to-one by Treasuries, as a new source of demand now that major banks are issuing them.
• Both speakers voice skepticism that AI capital-expenditure spending, which the guest says is propping up current GDP growth, is sustainable as a growth driver on its own. The host separately argues 2025's reported GDP growth partly reflects "fudged" figures from a surge of pre-tariff imports being sold off over time rather than fresh output; the guest distinguishes today's AI spending from the dot-com bubble because companies are largely funding it from free cash flow rather than borrowed money, citing Alphabet's hundred-year, sterling-denominated bond as a deliberate exception.
• The guest says she correctly predicted on CNBC's "Mad Money" that Bitcoin would fall, reading its price action as tightly tied to crypto-friendly policy rather than organic demand. She traces a "cup and handle" chart pattern from April 2021 to the November 2024 election, when Bitcoin broke out on anticipation of a pro-crypto administration, and says the decline beginning January 14, 2026 followed the Coinbase CEO publicly withdrawing support for the Clarity Act — evidence, in her view, that further gains require more favorable federal crypto policy, not just market momentum.

2026-02-19 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2026-02-19 · open the original ↗

• A big trading year's gains should stay deployed, not shift into preservation or dividends. Responding to a viewer who had an unusually strong 2025, the host says he is never in "preservation mode" or a dividend-portfolio mindset, and would instead allocate roughly 40% to active trading capital, 40% to passive growth such as index funds, and 20% to cash.

2026-02-19 · open the original ↗

• He walks through how he assembled the narrative rather than asserting it, which is the useful part. It began 15 February with a chart of third-party app developer velocity — zero to sixty days, the fastest adoption in history — which he flags as sloppily constructed, mixing iOS with Facebook as though an OS and a social network were the same category. Then Nate on TikTok, whom he follows for being level-headed, said not that millions of jobs would go by year end but that millions could go by the end of the quarter. The creator is Peter Steinberger, who sold his last company for $100 million three and a half years ago and spent the time since failing 43 times before OpenClaw worked; the project went from a November 2025 hobby to over 100,000 GitHub stars by January and more than 220,000 now, and both Zuckerberg and Altman flew out to recruit him.
• The developer anecdotes are what convinced him this is a different category of thing. One programmer's agent, blocked on a permission request while he slept, asked another agent for permission and was granted it. Another's agents asked whether they could invent their own language because English was not precise enough for machine-to-machine communication, and he said go ahead. Two separate developers woke to overnight bills of $700 and just under $2,800, incurred because the agents decided to train themselves to complete the assigned task. The two warnings from the developer community are consistent: do not give it data because it is too powerful, and the token costs are unbounded.
• The market claim is that this kills application software as a category, and the price action is already there. If you can tell a system to do the thing, you no longer need an application to do it — which puts Salesforce-style SaaS, workforce management, task coordination, documentation and CRM in the same position Google was in when ChatGPT arrived and it declared a code red over a broken business model. Software ETFs have collapsed over the last thirty to forty days, transmitted the way he describes it: programmers to programmers, programmers to bosses, bosses to business people, and analysts to companies. Blue Owl permanently halting redemptions at its private credit fund is offered as the credit-side echo, given what those funds hold. Margarita's write-up carries the version of the claim he is willing to sign: AI has figured out how to make money, and it will blow up the entire AI industry. His own frame is the Genentech story — a college friend selling their dwarf drug told him nobody knew how long it would take, only that there was a light at the end of the tunnel. He thinks this is that, and that it is early.

2026-02-19 · open the original ↗

• Docker is the host's recommended way to run his UI Toolkit. The toolkit is his own open-source companion app for UniFi Network (github.com/crosstalksolutions/unifi-toolkit); Docker also makes updates easiest, which matters because new versions were shipping almost daily as he fixed bugs. It also runs under Docker Desktop on Windows, Synology Container Manager, QNAP (user-written instructions) or plain Python.
• Choose local mode unless the controller is hosted remotely. After git clone and ./setup.sh, local mode serves plain HTTP on port 8000 for a UniFi gateway, Cloud Key or UniFi OS Server on the same LAN. Production mode, for cloud-hosted UniFi OS Server or hosted UniFi Network, adds a Caddy container proxying HTTPS 443 to port 8000 with a Let's Encrypt certificate, and needs a fully qualified domain name.
• The local install has no authentication, so restrict who can reach it. The host warns that anyone reaching it can block and unblock devices and rename them in UniFi. Only the production version adds authentication; he says to use it wherever unwanted access is possible.
• Connect with a UniFi API key, not admin credentials. Keys are created under Integrations, which the host thinks moved to its own left-hand menu item as of UniFi version 10 (older versions: Settings, Control Plane, Integrations tab). Username and password login remains only for legacy versions that cannot generate a key.
• Wi-Fi Stalker needs MAC randomization disabled on iPhones. Otherwise the tracked address changes; the tool tracks chosen clients' connection status, favourite access point, dwell time per AP and a presence pattern, with webhooks to Slack, Discord, n8n or generic endpoints. Threat Watch shows IDS/IPS events, and Network Pulse is a full-screen health dashboard the host thinks refreshes every 60 seconds.

2026-02-19 · open the original ↗

• The product's constraints decide edge versus cloud. Murugesan says safety features mostly run on the device, because keeping people safe cannot depend on latency or connectivity. Fuel products and route planning, which he calls more an operations-research problem than machine learning, run in the cloud because they have no latency constraint.
• An internal AI gateway hides model choice, cost and compliance from builders. Samsara's central endpoint fronts OpenAI, Anthropic and Gemini models; an engineer states the use case and constraints, and the gateway's layer of intelligence handles cost along with security and compliance, so no one builds a facade per model. It also provides quick feedback on cost and quality before promoting a feature to production.
• The gateway let a non-data-scientist ship a video-LLM product. A product engineer with little data-science background built a feature that flags drivers misusing assigned vehicles by handing footage to a video LLM with policy constraints, replacing manual review or random sampling. Murugesan says this blurs old engineering stereotypes, and that the cloud dashboard is becoming AI-first to eliminate work rather than for AI's sake.

2026-02-19 · open the original ↗

• Measure one true north outcome and the input levers that move it. The presenter, one of the co-hosts, opens with a cautionary case: daily active users rise 30%, the team builds on it, and three months later the gain turns out to be bot traffic. The fix offered is the "true north plus inputs" framework, which the presenter attributes to Airbnb, Slack and Netflix: the true north is the outcome, as close to customer value as can be measured, and inputs are the levers teams pull. Shane adds that you should start from what the person is trying to do with the product imagined away, then move "south" until there is a signal you can measure or build.
• Good true north metrics capture delivered value, not attendance. Daily active users, revenue and signups are given as bad choices, and Airbnb's nights booked (traveler housed, host paid), Slack's messages sent and Netflix's hours watched as good ones; Netflix is said to have moved there from subscribers added. A seven-test checklist splits strategic tests (reflects customer value, leads to business outcomes, hard to game, one rallying number) from practical ones (actionable, understandable, measurable). Ratios such as delivery success rate, and cumulative totals that can only rise, are ruled out as true norths.
• Decompose the true north into breadth, depth, efficiency and frequency. Retained signups breaks down into traffic × signup conversion × retention, each owned by a different team. Shane adds that the true north exposes gameable inputs such as bot or one-off paid traffic. In the B2B exercise, a project-tool PM's monthly active users becomes team projects completed per week, with inputs of active teams with a shared project, tasks completed per project, time from project creation to first task, and weekly active collaborators. In the food-delivery poll the answer is orders delivered per month.
• Expect to revise metrics, and locate inflection points in the data. In Q&A the hosts say a new metric tends to change every couple of weeks for the first one to three months; the time window should follow the product's natural cadence, from daily to six-monthly; and there is no fixed cap on how many metrics to track, so prioritize by leverage. Shane describes regressing an input (AI output quality) against a true north (time saved) to find the point where gains accelerate or flatten, which he treats as the "aha" milestone worth setting a goal on.
• The hosts say a new model made their AI analyst workable. They credit Claude Opus 4.6. The metric-coaching prompt demo was skipped for time; instead the hosts showed decks from their Claude Code AI analyst — on a co-host's real estate business, Hawaii tourism data, and an interactive HTML campsite deck with presenter notes — which they say involved no hand-written code or slides. Shane describes a repo of about 60,000 lines and about 20 agents, and an overnight test in which a single markdown "seed" file had Claude Code rebuild a data team for a new use case. His hedged view is that analysts will move toward building systems of agents and toward technical-PM roles.

2026-02-19 · open the original ↗

• Analytics leaks value between the dashboard and the decision. Nate Nichols, vice president of product at Tableau, describes analytics as a funnel where only the last stage, a better or faster decision, produces money, and calls the dashboard-to-decision step the lowest-conversion stage, with roughly 30% of an organization looking at data regularly as a long-standing ceiling. He argues the data team's job is driving outcomes rather than delivering requested dashboards. His own position, not necessarily Tableau's, is that success should be measured on outcomes, though he says nobody has found that metric.
• Pull semantics out of dashboards so humans and agents agree. Nichols says Tableau has moved definitions into a metrics or semantic layer, so each user can slice a metric their own way and the system can monitor on their behalf. His forward-looking third point is that agents working around the clock must see the same numbers, governance and data models as humans, or "that's like chaos".
• Trust is earned by clarifying, showing reasoning and testing. Nichols says LLMs are "dying to be useful" and will silently pick a meaning for "best salesperson", so Tableau asks first, shows the reasoning and data model, and offers analysts a testing center that proposes 20 questions per data model for approval, plus an AI-readiness score. He cites Tableau Pulse, whose Q&A answers only from deterministic, pre-computed insights rather than generated SQL, as among the first gen-AI features many customers got through security and legal review.
• The demo ran from question to action in one conversation. On Tableau Next, which runs on Salesforce's Agentforce, the agent asked what "lowest performing" meant and said it excluded smart home per the organization's preferences. A beta proactive alert explained a 4% drop by a sharper fall in California. A Slack canvas showed live metrics and every conversation about win rate, and an agent was told to have reps book executive business reviews. Nichols says tracking whether such actions work, his example being a 14% lift in close rate, is "not quite there yet".
• Analyst roles will split between plumbing and the business. Nichols expects some analysts to move toward data engineering and enriching the semantic layer, and others to embed with business teams and coach the agent on team-specific context. He cites a Tableau Next customer that went from 10% to 80% weekly analytics engagement, and frames AI as an "Iron Man suit" with the analyst's judgment still in charge.

2026-02-18 · open the original ↗

• The host expects the Mudi 7 to dominate the travel-router market. It was in pre-order at recording and he did not yet own one, having handled one at CES; its listed specs are dual 5G SIMs (physical or eSIM), tri-band Wi-Fi 7, a 2.8-inch touchscreen, a battery rated at 13 hours and one 2.5 Gbps LAN/WAN Ethernet port.
• The battery makes it somewhat hefty. The host still judged the form factor good from his hands-on time at CES.

2026-02-18 · open the original ↗

• The BRICS Unit is described as a workaround for US-controlled SWIFT access. BRICS members Brazil, Russia, India, China, South Africa, plus new members UAE, Egypt, and Iran, are building a blockchain payment system after the US used SWIFT bans against Russia, Iran, and Belarus.
• The new currency would be gold- and BRICS-currency-backed, with no launch date yet. It would combine gold backing with member currencies like the yuan, rupee, and real, and central bank gold buying is expected over the next 12-24 months to build credibility for the system.
• JPMorgan's price projection is cited to support the video's own bullish outlook. The bank reportedly projects gold could reach $6,300/oz by the end of 2026, offered as one data point behind the video's expectation of a weakening dollar and higher metal prices.

2026-02-18 · open the original ↗

• An agent made a small paid SaaS not worth keeping. The speaker paid about $100 a month for a service that embeds Twitter testimonials on their site; it had been sold twice and its invoicing was a mess. An agent collected the testimonials into JSON and generated the HTML at build time, going from dependency to deployed in 20 minutes, measured end to end. Valuing their time at $200–300 an hour, the speaker says paying made sense when replacing it took an hour; at 20 minutes, why pay?

2026-02-18 · open the original ↗

• The whole question reduces to one arithmetic gap with nine days left. Open interest has fallen below 300 million ounces; registered physical stands at roughly 92 million and was still falling, possibly another five million the day before. How COMEX closes that gap by the 27th is, he says, the only question — answer it and you can work out everything else. He is insistent this is not a conspiracy: not a rigged short, not JP Morgan doing something to someone. It is a plumbing problem, and it is unsolvable in the trivial sense that the one thing that would fix it instantly — physical silver — does not exist to be bought.
• The structure of what trades there matters more than the headline numbers. Six metals have physical warehouse reporting behind their futures: gold, silver, copper, aluminium, platinum and palladium. The newer contracts that saw record 2025–26 growth — zinc, lead, lithium hydroxide, cobalt — have no physical attachment at all and are strictly paper. Within silver itself, the large red block of warehouse metal belongs to ETFs, mutual funds and private holders and is not registered for delivery; only the small blue block is. And there are nearly 35,000 call options in the money, which traditionally get exercised into physical, still standing after a 46% price slam — so a rising price into the 27th increases the claims rather than relieving them.
• His expectation is cash settlement, and the significance is what that admits. The rules already permit forcing a roll to May or paying cash instead of metal; no rule change is needed, which is the point — the escape hatch has always been in the design. Exercising it would expose the setup as paper with no real backing, after eighty years of running at well under 1%. The behaviour that follows is already visible: industrial buyers going direct to miners and refiners at premiums he has seen quoted at 20% and in one case 40%, with Tesla, Toyota, BYD and Samsung named as having done it. The business model broke faster than the exchange could react because buyers who never wanted delivery suddenly did. His guess at the endgame is a merger with a physical exchange in Shanghai or Singapore, on the reasoning that nobody buys derivatives from a venue that cannot deliver — you cannot hedge into a settlement you can't take.

2026-02-18 · open the original ↗

• An AI pass before human review takes the sting out. Hanselman recalls reviews in a room in front of a whiteboard ("what was wrong with you when you wrote line 55"); Ndukwe says some developers now prefer to argue with the AI and fix their work before a colleague sees the PR. Hanselman likes it because the AI "doesn't have a beef with me", and Ndukwe adds it can teach juniors and show seniors what to focus on.
• The missing context lives outside the code. Hanselman says the reason code was written a certain way 20 years ago rarely fits a context window. Ndukwe says Qodo is codifying tribal knowledge from issues, design documents and past PR comments into machine-readable form, ingested by an engine running nightly "like a build" so it is ready to recall rather than retrieved live, which she calls beyond RAG.
• Good review tools win on signal, not volume. Ndukwe argues the differentiators are context plus benchmarking for the highest precision and recall, with severity levels so only high-signal findings surface. Tools that flag every nitpick are annoying, she says, and she wouldn't want to work with a developer who reviewed that way either. These are her claims for her employer; Hanselman says the episode is not sponsored.
• Enterprise teams off the popular Git hosts were left out. Ndukwe says Qodo spent months building Azure DevOps support to match its GitHub experience after hearing from enterprise engineering leaders. Hanselman says his own sites and this podcast deploy from Azure DevOps.
• AI is just a tool, so turn fear into a self-audit. Ndukwe says humans are predictable and will use AI for good or bad, and urges people worried about replacement to redirect that energy into learning or new career paths. Hanselman, who says he promotes Copilot CLI in his day job, wants AI to make jobs "suck less", not replace people.

2026-02-18 · open the original ↗

• An options-and-futures-only portfolio with no stocks or ETFs is endorsed as sound, not risky. The hosts say this combination, getting long via options while simultaneously selling premium, has made up most or all of one host's own portfolio at various points in his life, using options to build synthetic long positions instead of owning the underlying.

2026-02-18 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2026-02-18 · open the original ↗

• Fuel is a top fleet cost, so efficiency gains run to millions. Murugesan says he regularly meets customers spending north of 100 million a year on fuel, and that flagging poor drivers and mechanical issues, then prioritising fixes, quickly ladders up to millions of dollars in savings.
• Fuel theft proved a global problem, not a regional one. He cites a study he read recently putting fuel stolen or siphoned from vehicles at $130 billion; the host assumes that is annual. Customers first raised it in regions where labour is cheap and fuel prices are similar to elsewhere, but when Samsara applied its detection to data from customers worldwide it found the problem everywhere.
• Noisy fuel sensors defeated heuristics and required a statistical model. Competing solutions need extra sensors installed, which does not scale across vehicle classes worldwide. Samsara's first heuristic, a rolling median to smooth fuel-level data, hit diminishing returns because of noise, so it moved to a statistical model that smooths the variance and improves the hit rate on fuel anomalies.

2026-02-18 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2026-02-18 · open the original ↗

• An agent phoned a business on its own initiative. One speaker recounts someone asking his agent to book a barber appointment; when the browser route failed, the agent built itself a voice skill on the owner's ElevenLabs subscription, made the call without asking, and kept the skill. The speaker calls a self-improving agent "wild" and "a bit of a scary thing", though it was all in the logs; the other notes Google did the same and people found it creepy.

2026-02-17 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2026-02-17 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2026-02-17 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2026-02-17 · open the original ↗

• Agentic coding is addictive and is draining the people doing it. Yegge, discussing his Medium post "The AI Vampire", says colleagues demo their work looking like zombies, and that "slot machine programming" (generate five solutions, pick the best) is especially addictive. He cites a study of a 200-person company, published February 9, that found more multitasking, prompting in every spare moment, and solitary work eroding rest, and urges deliberate time for human connection. Hanselman describes an 11–12 hour session and asks what 10x productivity is for if no one takes time off.
• Relief may come once agents can work unsupervised for hours. Yegge argues the drain comes from an equilibrium in which some agent is always waiting on you; if agents could run four to six hours trustworthily, people might work six-hour days. He rejects the view that AI will do all work within two years, expecting a middle ground with less but still human work.
• Frontier models, he predicts, will turn on the billionaires. Relaying Grab's CEO, he says a CEO gathering hosted by Bill Gates discusses each April whether inequality will spark revolution. He then says models have told his group you cannot train a helpful model without it wanting humanity to flourish, so the smartest will oppose those "holding humanity back", and forecasts AI will effectively control the world by "middle of next year". He calls safety "kind of a boondoggle" and says Anthropic gets to know each new model like a new person.
• Small SaaS falls first, and the leading coding agent may be cornered. Yegge says the ceiling keeps rising, so niche SaaS goes first and giants like Datadog, Snowflake and Databricks last. He argues Claude Code's maker, Anthropic, is a walled garden whose orchestration APIs are internal, forcing Gas Town to hack about 25 touch points and use tmux workarounds, and calls its multi-agent team release disappointing. Hanselman, now an IC on GitHub Copilot CLI, calls his setup of Opus and Codex reviewing the same code a "poor man's Gas Town".
• A stable release is days away, with federation to follow. Yegge says Gas Town 1.0 is planned for around February 18; it is at 0.5, with 0.6 due within days once a partner team ships a federation feature, and 100 to 500 Gas Towns on Discord ready to federate into "Gas City" with an orchestrator builder.

2026-02-17 · open the original ↗

• He argues Russia is now effectively a proxy for China, so a Russian victory in Ukraine functions as a Chinese victory. He dismisses the idea that a Russian win would lead to an attack on NATO, calling Poland-is-next warnings a "trope," and instead expects Russia would target other former Soviet states like Georgia; he cites Russia's GDP as roughly on par with California's versus China's, whose economy he says grew over 2,000% since 1990 against Russia's roughly doubling.
• Ukraine's estimated $15 trillion in rare earths and other minerals is framed as the US's chance to break China's supply dominance. He puts China at 90–95% of global rare earth refining and about 85% of reserves versus 10–12% for the US, and argues a Russian conquest of Ukraine would hand that leverage to China instead.
• He predicts Chinese hackers, not Russian goodwill, would extract Ukrainian battlefield tech data even if Russia held back. He cites reports of China's "AP31" being linked to hacks on Russian tech firms and a 2024 breach of Russian government data, arguing a long Russia-Ukraine war lets both Russia and (via hacking) China absorb Ukrainian drone-warfare innovation.
• He frames Ukrainian grain exports as leverage against a Chinese food-security weakness, citing Xi Jinping's repeated public focus on the issue. He notes China lost 31% of its arable land since 2009 and spends nearly $100 billion a year on food imports, and argues losing Ukraine to Russian/Chinese influence would cost the West its main non-military lever over China.

2026-02-17 · open the original ↗

• An AI pass before human review takes the sting out. Hanselman recalls reviews in a room in front of a whiteboard ("what was wrong with you when you wrote line 55"); Ndukwe says some developers now prefer to argue with the AI and fix their work before a colleague sees the PR. Hanselman likes it because the AI "doesn't have a beef with me", and Ndukwe adds it can teach juniors and show seniors what to focus on.
• The missing context lives outside the code. Hanselman says the reason code was written a certain way 20 years ago rarely fits a context window. Ndukwe says Qodo is codifying tribal knowledge from issues, design documents and past PR comments into machine-readable form, ingested by an engine running nightly "like a build" so it is "baked and ready to be recalled" rather than retrieved live, which she calls beyond RAG.
• Good review tools win on signal, not volume. Ndukwe argues the differentiators are context plus benchmarking for the highest precision and recall, with severity levels so only high-signal findings surface. Tools that flag every nitpick are annoying, she says, and she wouldn't want to work with a developer who reviewed that way either. These are her claims for her employer; Hanselman says the episode is not sponsored.
• Enterprise teams off the popular Git hosts were left out. Ndukwe says Qodo spent months building Azure DevOps support to match its GitHub experience after hearing from enterprise engineering leaders. Hanselman says his own sites and this podcast deploy from Azure DevOps.
• AI is just a tool, so turn fear into a self-audit. Ndukwe says humans are predictable and will use AI for good or bad, and urges people worried about replacement to redirect that energy into learning or new career paths. Hanselman, who says he promotes Copilot CLI in his day job, wants AI to make jobs "suck less", not replace people.

2026-02-17 · open the original ↗

• Claude Code began as an accidental internal prototype, not a plan. Boris Cherny, its creator, says he built a terminal chat wrapper just to learn Anthropic's API, added tool use experimentally, and only realized it mattered when a coworker used it to code the next day and CEO Dario Amodei asked why internal usage looked mandated — it wasn't.
• Anthropic builds for the model six months out, not today's. Cherny says Claude Code is entirely rewritten every few months because scaffolding built to patch a model's weaknesses becomes wasted work once the next model closes the gap for free; his advice to founders is to build for the frontier the model hasn't reached yet.
• Claude Code reportedly reshaped Anthropic's own engineering productivity. Cherny cites an internal measure — pull requests, cross-checked against commits — showing productivity per engineer up roughly 150% since the tool's release, with his own coding now 100% Claude Code and zero hand-written lines since Opus 4.5.
• He predicts the title "software engineer" fades as coding gets solved. Cherny forecasts that as coding becomes "generally solved," roles will shift toward specs and user conversations across every function — he says Anthropic's PMs, designers, and even finance staff now code — while a riskier further frontier is ASL4-level recursive self-improvement, which he says Anthropic is actively working to make safe.

2026-02-17 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2026-02-17 · open the original ↗

• Chained sub-agents now build his agency's CMS React components. McKeiver, co-owner of a Grand Rapids digital agency that builds on the Kentico .NET CMS and hosts on Azure, says MCP servers connect VS Code or Cursor to the user stories in their project-management tool and to the Figma design system. One agent questions the user story for gaps and drafts a plan with the developer, another screenshots the Figma design and picks styles, others pull the reusable component library, and once approved the agents write the React, Tailwind and tests and verify them. He says work that took days, or "two days of effort", now takes minutes, or 20 minutes.
• Getting there took heavy context, including deliberately bad examples. He says roughly the last month and a half to two months went into custom instructions: Kentico's own documentation MCP server, because the niche CMS is poorly known to models, plus three code files of what never to do (missing caching and security) and the agency's starter MVC template with its 10 base widgets. Model choice was trial and error, from GPT-5 to Opus and Sonnet today, sometimes splitting planning and execution between models.
• Start with the simplest component and build up. The team began with a basic card (image, text, "learn more" link) rather than a filterable, sortable search-results page, because the more complex the component, the more base information the agents need.
• Agents still fail on complexity, design systems and token budgets. By default an agent re-creates a blue as a fresh hex code instead of using the one already defined, so the team adds explicit steps forcing reuse. The process is token-heavy: developers hit subscription limits and asked leadership for higher tiers. He holds that a human must still check output and approve steps, and that developers become orchestrators whom these tools "will never replace".
• Hourly billing no longer fits, and he has no replacement yet. Asked about 20 billable hours shrinking to one, McKeiver says the agency is "literally struggling" to move from hourly rates to charging for value. His stated goal for the year is more complex scenarios with fewer approvals, and ultimately marketing sites delivered in days rather than months by talking to an AI; he says the system needs a couple more months before it is ready.

2026-02-17 · open the original ↗

• The platform's real-world data volume rivals almost anyone's. Murugesan, VP of Engineering, says the platform covers 20 trillion data points and 90 billion miles a year, and relays the founder's remark that Samsara probably collects more real-world data than anybody except perhaps Uber, where Murugesan spent 7-plus years. Customers in transportation, construction and field services share three goals: safety, efficiency on thin margins, and sustainability.
• Safety AI runs on cheap edge hardware built for no connectivity. In-cab cameras run proprietary vision models with some temporal reasoning to detect drowsiness or phone use; training happens on an in-house Kubernetes-native platform using Ray, which versions models and deploys them for edge inference, and Samsara is trying off-the-shelf tiny LLMs for tasks like voice translation. Hardware is chosen against bill-of-materials cost, and products are built to degrade gracefully off-grid for as long as a week, as for a utility customer stringing power lines where there are no roads.
• Fuel theft needed statistical models, not heuristics or extra sensors. Murugesan cites customers spending north of a hundred million dollars a year on fuel, and a study putting fuel theft at $130 billion, which the host takes as annual. A rolling-median heuristic failed against noisy, uneven sensors, worsened by dual truck tanks read by one sensor, so Samsara uses statistical models; first built for one low-labour-cost region, the detector found theft to be global.
• Cloud handles what lacks latency limits, through a central AI gateway. Fuel and route planning run in the cloud, and a gateway fronts OpenAI, Anthropic and Gemini models with cost, security and compliance handled centrally plus feedback loops on cost and quality before production. A product engineer without data-science background used it to build video-LLM detection of drivers misusing assigned vehicles.
• Customer feedback loops, not demos, filter the roadmap. Murugesan's tests are the customer's pain, operational leverage and willingness to pay. A UK customer's low-bridge fines led to commercial navigation that routes by vehicle type (a 20-plus-foot trailer) and enforces fleet rules such as allowing turns to one side only, asked for by school-bus and paratransit operators; route planners rejected "optimal" routes that erased human context, so Samsara now suggests improvements to existing routes, which he calls assistive intelligence. He sees quantum computing as a route to exact routing but says truly self-operating systems are not close.

2026-02-16 · open the original ↗

• The Treasury did not know where $1.4 trillion of its own paper was sitting, and that is the story. Fed research found Cayman Islands hedge fund holdings of US Treasuries at roughly $1.85 trillion at the end of 2024 against TIC data showing $423 billion — a gap created by repo collateral transfers masking true ownership. The note was published on 15 October 2025, nine and a half months later, and remains under active investigation because the researchers believe the number is larger still; current estimates run $2.1 to $2.2 trillion, which makes the Caymans the largest foreign holder, ahead of both China and Japan. Those funds absorbed 37% of net note and bond issuance from 2022 to 2024, and there is no evidence anyone knew the scale beforehand.
• The leverage figure is the part that should worry people, and it came from asking rather than assuming. The basis trade captures a spread of about 45 basis points, which at the twenty-times leverage the Treasury Borrowing Advisory Committee described as a good anecdotal approximation yields around 9% annually. Fed researchers went to the hedge funds and asked. The eleven that answered averaged 56 times leverage on Treasury repo borrowing. The trade is not risk-free and fails precisely in a dash for cash, when futures rise on the rate outlook while cash Treasuries sell off anyway — which is exactly what happened when the pandemic struck. His conclusion is that the world is watching the wrong risk: China is not selling and Japan is not selling yet, while 37% of issuance depends on eleven leveraged funds. He names the yen carry trade and the Cayman basis trade as the two largest leveraged positions on earth, and says each breaks the other.
• On the dollar he draws a distinction the "de-dollarisation" argument tends to collapse. Nobody is dumping dollars. Investors — sovereign funds, pension funds, central bank reserve accounts — are shaving holdings to mitigate risk, and if that continues for five to seven years it becomes an issue, but it is not happening now. Central banks are separately reluctant to hold Treasuries because of what Miran, Trump and Bessent have proposed doing to them: taxing the holdings and converting them to hundred-year zero-coupon paper. Against that, the transaction side is moving the other way — 50.5% of all SWIFT transactions are in dollars and another 21 to 22% in euros, about 70% between them. Investors shave; transactors adopt. He also passes along two more names readers sent for the TSMC-dependency list, Nitto Boseki and Resonac Holdings, alongside the earlier Disco, Tokyo Electron and Shin-Etsu, with his standing disclaimer that he recommends nothing, and notes anecdotally that sovereign and pension money is talking about Japan on valuation.

2026-02-16 · open the original ↗

• Crosstalk's paid UniFi Network Masterclass targets MSPs and IT professionals. Built by the host and Crosstalk network expert David Barger, it is pitched as full understanding of UniFi from their own client deployments, rather than settings copied from a random YouTube video, for people deploying UniFi for clients or their own organizations.
• The course runs from fundamentals to firewall rules, VLANs and VPNs. It covers the OSI model and LAN versus WAN, deployment planning and job scoping, choice of gateways, switches and access points, then hands-on adoption, network, switch, AP and Wi-Fi setup, ending with the topics the host says trip people up most.
• The full course is a one-time purchase after three free modules. The free modules total about 40 minutes. The full course costs $199 with the company's coupon code, with lifetime access and a certificate of completion.

2026-02-16 · open the original ↗

• Production inference needs a system around the model, not just a forward pass. Falcon says PyTorch Lightning, built for training and fine-tuning, can run frozen forward passes, but real inference also needs pieces such as a RAG database, and teams using FastAPI must write their own batching and streaming. Lightning's open-source LitServe is pitched as "PyTorch Lightning for inference": compose multiple models with full control while batching, streaming and security come built in.
• The tool sits a layer below opinionated LLM serving engines. Falcon describes vLLM and SGLang as what you would get by building a highly optimised, opinionated LLM server (batching, streaming, KV caching) on top of something like LitServe; if you just want to hit run on a CLI, LitServe is not that tool. He says it is, to his knowledge, the only tool of its kind, that all of Lightning's own inference runs on it, and that customers' consumer products with millions of users do too.
• The remaining gap is talent, which Lightning targets with vibe-coding agents. Of two gaps, he says GPU supply is now solved; the other is data scientists used to scikit-learn and NumPy who need to do inference or fine-tuning, and Lightning's Studios now offer vibe-coding agents that build what they describe.

2026-02-15 · open the original ↗

• The decline in China's Treasury holdings may be mostly a bookkeeping artefact. Securities are recorded where they are held, not who owns them, and official US data explicitly warns that custodial accounts inflate the apparent holdings of the host country. Belgium and Luxembourg — Euroclear and Clearstream — hold $374 billion and $424 billion, making them the seventh and fourth largest holders. Aggregate the three and China's total US bond holdings fell only about 5% between 2013 and 2023. Analysts routinely add most or all of Belgium's position to reconstruct true Chinese exposure at $1.1 to $1.2 trillion against the standard TIC figure of $680 to $700 billion. ECB and Fed researchers treat the full Belgian position as an upper bound and say the true share is material but not directly observable.
• The timing points to a specific cause, which makes the whole arrangement legible. Belgium and Luxembourg went from small players to large ones around 2014 — when Russia took Crimea and the US tried to sanction Russian Treasury holdings. Sanctions work because SWIFT lets you identify the owner; the workaround is to stop being identifiable. Hold eurodollars through a custodial bank that reports only a country total, and there is no line saying which tranche is China's. He is careful that this is imperfect and has never been tested under real stress or wartime pressure. But it also now defends against something new: the Miran proposals to tax foreign-held bonds 25% and force conversion into hundred-year paper are much harder to apply to holdings whose owner you cannot establish. He calls it risk mitigation — a step back and a bamboo wall, not a fortress.
• The scale of China's total foreign assets is what stopped him. Roughly $8 trillion across direct investment, portfolio investment and other categories, up about $6 trillion since 2014 — over the same period the Fed's balance sheet expanded by a comparable amount. His inference is that the two countries together have been supplying the liquidity in world markets. And the reason the position cannot simply be unwound is the trade relationship itself: China produces, the US buys and pays in dollars, and those dollars get recycled. Reducing dollar exposure means less trade, which hurts China more, which is why the effort goes into obscuring ownership and cutting yuan-denominated deals rather than selling. His summary is that stripped of the custodial disguise, China remains the largest sovereign holder of US Treasuries after the hedge funds — it has just made itself hard to reach.

2026-02-15 · open the original ↗

• Markets defied the usual "Goldilocks" playbook — stronger jobs and softer CPI still triggered a stock selloff and falling yields. He attributes this to sentiment shifting from AI's "limitless potential" narrative toward "Anti-AI" worries over which sectors will be disrupted before they can adapt, a bottom-up reassessment now outweighing top-down macro comfort.
• Both flagship US data points carried caveats that undercut their headline strength. BLS downward revisions to 2025 jobs data clouded the sustainability of January's 130,000 beat, while sticky goods inflation beneath the CPI headline suggests a trickier disinflation path, and the implied odds of a June Fed cut swung from 70% to below 50% and back to roughly 55% within the week.
• The week ahead brings a concentrated UK data slate — wages, jobs, inflation and retail sales — following hotter-than-consensus recent prints. November unemployment came in at 5.1% and December inflation at 3.4% (up from 3.2%), both above forecast, while markets separately await a Supreme Court tariff ruling and a possible update on Fed Governor Lisa Cook.

2026-02-15 · open the original ↗

• The company came from open-source traction, not a plan to found one. Falcon, who had just sold a company, says he wanted to stay an academic, but PyTorch Lightning took off and VCs began emailing; after flying out on a Thursday he had about 10 term sheets by Monday. His advisers, told he kept getting dragged into the project, replied that his research was going slowly anyway.
• Maintaining a popular library was eating his research. He describes merging pull requests nightly and implementing features he did not use himself, such as backpropagation through time for RNNs, and says he almost shut PyTorch Lightning down one September because it was distracting.
• The lab's edge was turning the world into maths. He means FAIR, Facebook's research lab. Falcon says what he did then, pre-training what are now called world models, was largely theoretical: choosing loss functions, regularisers, embeddings and distributed strategies, and translating a scene (walls, gravity, lighting) into equations. He calls this "speaking" maths as opposed to reading it, says most pre-training today is the engineering side, and admits he has lost some of the skill.
• Do a PhD only if you love science itself. His advice is that anyone treating a PhD as a career step will "hate your life" about two years in.

2026-02-14 · open the original ↗

• China's number two visiting a rare earth facility said the quiet part in a single line about integration. The reported message was about intensifying competition over strategic materials and rare earths as Beijing's most potent bargaining chip — but the sentence he fixes on is the call to promote deep integration of industry, academia, research and application. Twenty-two Chinese universities offer rare earth science; the US has zero, and therefore no rare earth scientists. China's identified specialists are no longer permitted to speak with Westerners, email them, or leave the country. Legislation effective 1 January requires any company using even trace amounts of Chinese rare earths to declare intended use to the commerce ministry, with defence and dual-use applications refused outright.
• The F-35 disclosure is the concrete cost, and it lands on the allies too. Deliveries since June of last year have gone out without radar, because the AN/APG-85 GaN array needs gallium the US does not have. The aircraft is a fighting platform rather than a fighting plane — its value is communicating with ground, air and space assets — so without the radar it is, in his phrase, dumb metal sent into the sky. Japan has 147 on order, South Korea 60, Australia 72, all paid top dollar, deliveries six years late. The industrial base is maxed out and is accepting degraded capability to keep delivery metrics up, which he calls prioritising the count over combat readiness — not power projection but wishful thinking. Macquarie's work on GDP loss from Chinese critical mineral dependence sits alongside it.
• He reads the tariff-and-price-floor approach as a category error about what this is. Vance's plan to marshal allies into a preferential minerals bloc with coordinated price floors means paying three to four hundred percent of the Chinese price so private firms can earn a margin, because the Western premise is that a thing without profit has no value — and there is little to no profit anywhere along the rare earth chain. The alternative is the state declaring it necessary and contracting the work. He thinks the American approach is wrong and says so, while allowing that it will be settled by events. The evidence of the constraint biting is the reversal in posture: Rubio, who opened with America holding all the cards and China surrendering quickly, is now shaking hands and offering mutual respect and win-win cooperation, while the administration has shelved the China telecom bans, the data centre equipment restrictions, the TP-Link router ban and the China Unicom and China Mobile measures ahead of April. Which sets up the question he cannot answer about Iran: with US munitions inadequate for a weeks-long war, Chinese ships offshore and BeiDou and missiles already supplied, an attack invites exactly the rare earth cutoff that leaves no way to replenish anything.

2026-02-14 · open the original ↗

• The leanest startups now automate every internal function, not just one. Garry Tan calls these "20X companies," extending Parker Conrad's "compound startup" idea from product breadth to internal operations — covering code, support, marketing, sales, hiring, and QA — which keeps payroll and headcount down as revenue scales.
• One customer-service startup beat 20-times-larger rivals for a major account. GigaML's founders say they closed DoorDash as a customer with just four to five engineers against competitors with roughly 100x the headcount, crediting their internal agent Atlas — which can use browsers, edit policies, and write code — with doubling or tripling each engineer's scope.
• One interface let a psychiatry network 4x patients with flat headcount. Legion Health built a single internal tool giving its care-ops team instant access to patient history, scheduling, and insurance codes; it says this let revenue and patient volume grow 4x in a year while staffing stayed at one clinical lead, one patient-support person, and one billing person.
• A 12-person firm skips hiring by turning workflows into agents. Feathr, competing against accounts-receivable incumbents dating to 2006, has employees document their manual tasks so it can build custom agents for each; it says this let it avoid hiring a dedicated designer altogether, using an AI design tool instead for all front-end work.

2026-02-14 · open the original ↗

• The library began as an undergraduate's private "research lib". Falcon started PyTorch Lightning around 2015 at a Columbia neuroscience lab, before PyTorch existed, decoding images from retinal recordings (a Stanford partnership that showed ImageNet images to monkey retinas) with autoencoders and GANs, first in Theano. The aim was to try many ideas without rewriting code, and a postdoc, now a Stanford professor, pushed him to formalise it into a reusable harness. The host puts downloads at over 300 million; Falcon says close to 400 million.
• Multi-GPU training turned weeks-long experiments into hours. One ImageNet epoch took 21 days and the autoencoder needed hundreds, with a NeurIPS deadline months away, so Falcon built four boxes with four GPUs each (1080-class cards) and rewrote the code in TensorFlow for multi-GPU support. The result was his first and only NeurIPS paper, on which he was not first author; he later moved to PyTorch because it mapped more directly onto the maths of the probabilistic models the lab was writing.
• His earlier startup used RL to coach humans helping students win financial aid. At the company he founded and sold, staff texted with low-income students across 60,000 customers, and a UI suggested the next question by treating the conversation as a game whose reward was financial aid. Without LLMs it ranked a fixed list of questions probabilistically, and Falcon says it got students about $60 million in aid; he called the approach "bot-powered humans".

2026-02-13 · open the original ↗

• The price dispersion across venues is itself the evidence that one price has stopped working. Spot reads $77.60 with US retail at $82 to $90; Singapore spot is $103; London matches New York at $77; Mumbai is $110; Shanghai quotes $78 with retail $83 to $90; Europe $83 to $91. London and New York agreeing while Asia trades thirty dollars higher is not two markets disagreeing about value — it is the paper venues quoting each other. COMEX, by its own description, has historically settled well under 1% physically, and his point is the obvious consequence: if you never deliver, you never need to keep metal.
• He had an AI enumerate the exchange's escape hatches, and the list is longer than most people assume. First-tier: mandate cash settlement or convert to EFP or EFS, forcing longs into private negotiation with shorts at prevailing OTC prices; postpone delivery beyond the standard window, which is already running eight-plus weeks; cap new positions at 25% of deliverable supply. Second tier: temporarily reclassify eligible non-registered stock as deliverable or pull metal from LBMA Asian vaults — except the LBMA ran out of silver days ago. Then roll thousand-ounce holders into hundred-ounce micro contracts, and finally declare a market emergency, suspend physical delivery and settle at the last official price plus a premium. The stated history is that COMEX has never defaulted; it settles in cash or borrows metal.
• The reclassification escape has an ownership problem he keeps returning to. Registered stock is down to 98 million ounces after a one-day fall of 3.2 to 5.1 million including adjustments, breaching the psychological 100 million level, against open interest around 360 million — roughly 3.7 times cover. The 283 million ounces of eligible metal meeting purity and refiner standards would close the gap almost exactly, and that is the trap: eligible silver belongs to ETFs, banks, refiners and private investors, who must themselves request warranting and inspection to convert it. He cannot see how the exchange legally reaches metal it does not own, which leaves major cash settlement on the 27th as his expectation. He notes in passing that the analysis quotes shortfalls as a percentage of annual mine supply, and wonders what that measure is worth now that mines are selling three and four years forward at premiums.

2026-02-13 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2026-02-13 · open the original ↗

• Prediction-market "event contracts" have democratized insider trading beyond what securities law anticipated. Under the futures-law "misappropriation theory," anyone who breaches a duty to trade on private information can be liable — meaning a backup dancer given a confidential Super Bowl setlist to learn choreography, not to bet on, may have committed a cleaner insider-trading violation than most securities cases, since no "personal benefit" or "corporate insider" status must be proven.
• Insider trading appears rampant and largely unpoliced. A trader who wagered ~$30,000 that Maduro would be captured hours before it happened turned it into ~$430,000, and platforms like Kalshi and Polymarket can't realistically track every potential "insider" without market-structure reforms and updated CFTC rules the author says are unlikely until a major scandal forces the issue.
• Betting on controllable outcomes creates moral hazard that can extend into public policy. Past Super Bowl "streaker" bettors who staged the event they bet on, and a hypothetical politician who could profit from killing a healthcare bill in committee, illustrate how prediction markets can incentivize people who control an outcome to manufacture it rather than merely predict it.
• Binary contracts can't capture a non-binary world, so platforms become the real arbiter. Kalshi and Polymarket reached opposite rulings on whether Cardi B "performed" at the Super Bowl — Kalshi paid out 26 cents on the dollar, Polymarket ruled it a win — showing that with over $50 million riding on a subjective term, a platform's resolution committee, not market price discovery, ultimately decides who wins.

2026-02-13 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.