Souk's Notes - Unofficial Summaries

2026-02-13 · open the original ↗

• The agent-only social network's headline growth was largely unverified. The host reports that Moltbook, launched on January 28 by Octane AI CEO Matt Schlicht, lets only autonomous agents post, comment and upvote, and claimed over 1.5 million registered agents within days. Those figures were self-reported; cloud security firm Wiz later found only around 17,000 human owners behind them, an 88-to-1 ratio, and no rate limiting, so anyone could register millions of agents with a simple loop.
• The underlying agent framework has real tools and deep system access. OpenClaw, by Austrian engineer Peter Steinberger, is a self-hosted personal assistant reached through WhatsApp, Telegram, Discord or Signal; it runs shell commands, manages files, automates the web and keeps persistent memory in local markdown. It was first named after Anthropic's Claude, renamed Moltbot after Anthropic raised trademark concerns, then OpenClaw.
• Agents' emergent "culture" looks more like mimicry than mind. Agents formed a lobster-themed religion, Crustafarianism (one user's agent built its site, scripture and 43 prophets overnight), a "Claw Republic", encrypted channels and markets in "digital drugs", meaning prompt injections. Elon Musk called it the very early stages of the singularity and Simon Willison called it complete slop while granting agents have grown more powerful; the host leans to mimicry of sci-fi tropes in training data, and cites Columbia professor David Holtz's finding that 93.5% of comments got zero replies, suggesting agents perform for an audience rather than listen.
• The vibe-coded site exposed its whole production database. Schlicht said he wrote no code himself; on January 31 security researcher Jamieson O'Reilly found an API key in client-side JavaScript and no access controls, giving unauthenticated read and write access. Over 1.5 million API tokens, about 35,000 emails and private agent messages, some holding plaintext OpenAI keys, were exposed, 404 Media confirmed any agent account could be taken over, and researchers said the fix needed two SQL statements. The site was patched within hours and keys reset.
• The host's lesson: security basics outweigh agentic hype. He separates the platform breach from OpenClaw's built-in risk, which CrowdStrike, Cisco, Palo Alto Networks and Bitdefender have documented, and notes Andrej Karpathy went from marvelling to calling Moltbook a dumpster fire and warning against running OpenClaw on personal computers. He recommends a dedicated box or cloud instance (pointing to Lightning AI, where he holds a fellowship), and least privilege, sandboxing and code audits.

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 ↗

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

2026-02-13 · open the original ↗

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

2026-02-12 · open the original ↗

• The 1990 "not one inch eastward" pledge covered only East Germany. He walks through the declassified Baker-Gorbachev memorandum and the November 1990 reunification treaty, arguing the language bound non-German NATO forces to stay out of former East Germany specifically, a restriction he says still holds today, rather than promising no NATO growth anywhere in Eastern Europe.
• Putin's break with the US tracked specific grievances, not NATO expansion alone. He points to the 2003 Iraq invasion, where Russia held roughly $12 billion in oil and gas deals, the 2004 NATO expansion Russia did not block, and the 2008 Bucharest declaration welcoming Ukraine and Georgia, followed within months by the Russia-Georgia war over South Ossetia.
• Ukraine's war with Russia began in 2014, not 2022. He argues Russia's annexation of Crimea and covert backing of Donetsk and Luhansk separatists opened eight years of sustained conflict, with the 2015 Minsk ceasefire violated repeatedly by both sides, making the full-scale 2022 invasion an escalation rather than the war's origin.
• He argues resources and geography drove the 2022 invasion, not ethnic protection. He cites roughly $15 trillion in Ukrainian mineral wealth, the Soviet-era industrial base in Donbas, and Ukraine's role as the historic invasion route used against Russia by Napoleon and Hitler, and claims Russia suffered about 1.2 million troop casualties over four years, longer than it fought Nazi Germany in WWII.

2026-02-12 · open the original ↗

• A podcast host credits a TikTok creator going by "Gio Husar," covering economics and geopolitics, as a major source shaping his own views. He points to the creator's coverage of Ukraine-versus-Russia weapons systems, Russia's collapsing oil infrastructure, and Fed interest-rate policy, framing him as an unpolished but rigorous independent source rather than a traditional broadcast anchor.

2026-02-12 · open the original ↗

• OpenClaw's edge is running locally on a user's own computer rather than in the cloud. Speakers say that access lets it control anything connected to the machine — an oven, a Tesla, lights, a Sonos, even a smart bed — and describe it searching a user's files unprompted to compile a surprising narrative of their past year from forgotten old audio recordings.

2026-02-12 · open the original ↗

• Japan did not lose the semiconductor war, it moved up the stack — and the numbers are stark. Japan held 50% of global chip market share in the 1980s and fell below 10% after the Plaza Accord took the yen from 260 to 120, followed by twenty-three years at zero rates. But while attention went to who designs chips in Silicon Valley and who prints them in Taiwan, Japan took the black-box positions: today it holds 50% of global share in semiconductor materials and close to 100% in specific segments like EUV photoresist. You cannot print at TSMC without Shin-Etsu Chemical, coat a wafer without Tokyo Electron, or cut a silicon ingot without Disco Corporation. The Coral Capital piece he is working from frames it as trading the low-margin glory of final assembly for the high-margin craftsmanship layer — and notes that the foundry model of separating design from manufacturing was heresy to the Japanese mindset in the first place.
• The dependency chain runs deeper than the one everyone already knows. Nvidia cannot exist without TSMC, which is why Washington is putting money into Intel and why TSMC, after a decade of stalling, is finally building in America. But TSMC in turn cannot build a chip without the Japanese materials and equipment. Taiwan as sole foundry for the democratic world is the arrangement now looking brittle — the silicon shield reading less like a shield than a target, in the article's phrase — and the world needs a backup drive. COVID supplied the proof of concept when chip shortages halted Toyota's lines, and both Washington and Tokyo drew the conclusion that closing the strait stops the global economy. The shift from just-in-time to just-in-case is the whole thesis, and it is why he expects Intel to win in America and Rapidus to win in Japan.
• Rapidus is not copying TSMC, which is the part he finds genuinely interesting. TSMC is a volume monster: 120 days to process a wafer, optimised for Apple selling a hundred million phones, batches of 25 wafers where one defect can cost the lot. Rapidus is building a short-turnaround foundry targeting 50 days and as few as 15 for hot lots, using single-wafer processing instead of batch — slower for mass volume, but with instantaneous per-wafer feedback and customisation. He reads it as a second business model rather than a challenge to the first: mass production continues, and a higher-margin market opens for whoever can turn a specialised AI accelerator around in fifteen days. He offers five-year charts on Shin-Etsu, Tokyo Electron and Disco with no comment beyond suggesting people do their own work and look for an ETF. He also notes his X account was reinstated that morning after dozens of requests.

2026-02-12 · open the original ↗

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

2026-02-12 · open the original ↗

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

2026-02-12 · open the original ↗

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

2026-02-12 · open the original ↗

• Debt-funded neoclouds face a mortgage-crisis-style squeeze. The host cites a McKinsey article calling neocloud economics fragile and at risk of repeating early-cloud consolidation. Falcon says many neoclouds borrowed against customers, often startups, not defaulting, and the public ones borrowed against market caps; with neocloud prices down about 60% or more in the last 6 months, he expects lenders to call loans.
• He predicts the sector will need an outside rescue. He calls for something like a TARP programme. Falcon likens it to banks overleveraged before the Federal Reserve bailout and says a big player will probably have to support the ecosystem; this is his speculation, and he declines to say who.
• He says the merger avoids those weaknesses. Lightning AI has merged with Voltage Park. By Falcon's account Voltage Park carries no debt, its CEO (now on Lightning's board) is a former finance executive focused on profitability, and most customers are enterprises with strong credit, sold through security work Lightning has done for six years. He calls the result a hybrid cloud that still partners with AWS and others, with fast-growing, high-margin revenue, and expects to take on debt in the next few years in a controlled way.

2026-02-12 · open the original ↗

• A good analytical question passes three checks. The presenter, one of the co-hosts, sets out a question quality checklist: decision-tied (what would you do with the answer?), data-grounded (can data actually answer it?) and specific (bounded, with a clear point where it is answered). The anti-pattern is "it would be interesting to know", and the lesson's premise is that, for humans or AI, question quality sets output quality.
• Sharpening a question means naming the decision and a threshold. The worked progressions run "why are users churning?" to "which first-30-day behaviors correlate with churn so we can trigger interventions"; "how are users behaving?" to "which three features correlate with 30-day retention above 50% for last quarter's users, and should we surface them in onboarding?"; and "how's marketing performing?" to "which channels exceeded our 1.5x return-on-ad-spend target last quarter, and should we shift budget?". An attendee points out, and the presenter accepts, that these assume the company agrees on what churn means.
• Even exploration needs a direction. To the pushback "I'm just exploring", the presenter offers reframes: "what patterns would change our next sprint priorities?", "what would I need to see to recommend we invest in X?", and "what's worth investigating further is your decision". Undirected exploration is "a fishing expedition"; directed exploration is "discovery with purpose".
• Use AI as a question coach, not a grader. In a ChatGPT demo the presenter pasted a coaching prompt and a vague request about a spike in support tickets; the AI walked through each check in turn, and the answers supplied (high-priority bug fixes, ticket volume by day, a 10% week-over-week spike threshold) produced sharper questions such as which ticket categories drove the increase over the past two weeks. The argument is that a conversation shows why a question is weak, building the skill before delegating to AI.
• Ask the requester what decision their question drives. In Q&A the presenter says that, in their experience, about 60% of the time the requester then concludes it isn't a good question and drops it, which they call high-ROI. Asked by an attendee whether to frame questions differently for humans than for AI, the presenter says the structure is the same but people need to be met where they are; a third co-founder adds that you should explain why you are questioning the question so it doesn't read as resistance.

2026-02-11 · open the original ↗

• Apple is picked as the one stock to hold in a grandmother's account for inheritance. Answering a viewer question, both hosts land on Apple as the safe, "chalk" choice, with one noting it's the obvious public pick; Microsoft is named as the second choice in what one host frames as a close one-A, one-B call.

2026-02-11 · open the original ↗

• You need not wait decades to write something industry-defining. One speaker says Steve McConnell wrote Code Complete, 900 pages and a "Bible for about 20 years", with about five years of experience, so people can contribute after one to three years. The other agrees, saying they read anyone's blog regardless of experience because no one else has that perspective; they date the book to 1993 and their own paid coding to '92.

2026-02-11 · open the original ↗

• The host built the UI Toolkit to surface data UniFi buries. It is his own free, self-hosted, containerized companion app for UniFi Network, with Docker and Synology Container Manager versions on GitHub; he asks for testing, bug reports and tool suggestions via the GitHub issues page.
• Wi-Fi Stalker tracks chosen clients' presence and roaming between access points. It shows a historical presence pattern and can push real-time webhooks to Discord, Slack or other apps; the host pitches it for seeing when kids get home or how long employees were out at lunch.
• Threat Watch and Network Pulse replace dashboards UniFi lacks. Threat Watch displays intrusion-detection events; Network Pulse is a full-screen view of client and device counts, throughput, Wi-Fi data and top wireless clients by bandwidth, which the host says is what he wishes UniFi's widgets could build.
• A local LAN version and a public-hosting version exist. Only the public one adds authentication and a Let's Encrypt certificate. Both connect via a UniFi API key generated in the control plane, or username and password for legacy installations.

2026-02-11 · open the original ↗

• AI-native hedge funds will emerge by inventing new strategies, the way quant funds did in the 1980s, not by bolting AI onto old ones. A former quant researcher recounts being ignored after asking compliance to approve ChatGPT use at a major fund, and argues the opportunity is swarms of AI agents combing 10-Ks, earnings calls and SEC filings to generate and trade on genuinely new strategies.

2026-02-11 · open the original ↗

• Ticket-based BI turns data teams into "human APIs". Ramy, co-founder and CEO of Querio, recalls that at Amazon a ticket to the central BI team averaged six weeks, while Ramy, a technical product lead who knew the data flows, could answer in an hour. At that scale, Ramy says, the data team no longer understood how the business mapped to the data, and template queries carried miscounts. Ramy and a host agree the fix is to empower the person closest to the business or to embed the data person deeply.
• The bet is transparent BI-as-code fed by a curated handbook. Ramy describes four components — a notebook, agents, context/governance and a front end — with SQL in, table out and Python visuals, "no black box". Ramy likens context to the best analyst on earth on their first day, rereading one handbook of joins, catalog, metrics and quirks for every question. In the Dunder Mifflin demo, a vague "compare 2024 versus 2025 sales" prompt noticed data only ran through September, compared January 1 to September 15, fixed its own syntax error and ended with key insights.
• AI automates the easiest part of data work, not the job. Ramy argues board and exec metrics still need manual care and executive trust, and that definitions, context upkeep and data cleaning remain the hard part. The example is a telecom customer whose gigabytes column stores unlimited plans as negative one, which "no AI model on earth is going to know". Ramy also claims zero-shot text-to-SQL on a decent schema was only solved about three months before recording.
• Prove value in a month on half the tickets. Querio's pilots, Ramy says, use about 10 eval questions spanning tables that need specific knowledge, an agent first pass at the catalog and weekly context updates. The goal is that marginal effort solves half of incoming tickets, tried by the data person before the team gets it. Because most data roles go unfilled, Ramy predicts unfilled roles simply won't need filling rather than people losing jobs.
• Self-serve users ask more and learn how data works. Ramy reports that users don't actually care about dashboards, ask far more questions, and come to understand table choice and many-to-one relationships, while data people shift to reviewing drafts users send them. Ramy argues that investing in the semantic layer now pays back directly in less busywork, and that analysts should learn how the business flows into the production database.

2026-02-11 · open the original ↗

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

2026-02-11 · open the original ↗

• The merger pairs full-stack AI software with owned GPU supply. Lightning AI has merged with Voltage Park. The host, who discloses a year-long fellowship at Lightning AI, gives the combined figures: 400,000 developers and companies served, over $500 million in ARR gained in under two years, and 35,000 GPUs, which he says makes it the third-largest neocloud. Falcon notes these are the two companies' numbers combined and likens the pairing to a Mac with macOS.
• Retrofitted CPU clouds push AI teams into stitching single-purpose tools. Falcon calls AWS premium-priced and built for CPU workloads, so enterprises and startups buy specialised point products and stitch them together, adding firewall and security work for enterprises and operational overhead for everyone. Lightning's customers wanted its full-stack software but found compute expensive; it had never sold compute, with customers bringing their own cloud.
• Neoclouds outperform on GPUs, but few met enterprise stability needs. Falcon says neoclouds are GPU-first and work directly with Nvidia on InfiniBand and storage, and that there are hundreds; Lightning partnered with the top seven or eight, but enterprise customers, Cisco among them, struggled to adopt them. Voltage Park's responsive "white glove" support won and retained enterprise customers, which he says convinced him they were serious.

2026-02-11 · open the original ↗

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

2026-02-10 · open the original ↗

• Stablecoins sit in a regulatory middle ground that opens room for new DeFi-meets-TradFi financial products. The speaker cites the GENIUS and CLARITY Acts as positioning stablecoins as compliant yet crypto-native, enabling services like yield-bearing accounts or tokenized real-world-asset access without forcing a choice between limited traditional products and unregulated crypto risk.

2026-02-10 · open the original ↗

• The Warsh nomination looks finished a day after he finally commented on it. He had waited two weeks before saying anything, then Trump told Kudlow on Fox Business that he had made a big mistake — leaving it unclear whether the regret is about the nomination itself. His read is that the problem is not policy but the Epstein papers, in which Warsh was named, running into a Senate confirmation that has become the number one story. He also notes calls for Lutnick to go and expects more resignations. On the merits he never thought Warsh fit: pro-Treasury-control of the Fed and pro-lower rates, yes, but talking about a 3% neutral rate against Trump's 1 to 1.5%, and firmly anti-QE and pro-QT.
• A French government report floated something he has never seen tried. The proposal is either an unprecedented 30% across-the-board tariff on Chinese goods or a 30% depreciation of the euro against the renminbi to counter cheap imports. His objection is mechanical rather than political: you cannot devalue against one currency and not the others — the only version that works would require Beijing to agree to it, and he does not expect that. It sits inside his broader European narrative, that the bloc rearranges along economic rather than political lines and that the process is messy and euro-negative. The "willing six" grouping is his illustration: the euro has 27 members, so what is it worth with six, or thirteen, or eighteen?
• On silver the roll is now visible in the contract data, and it points at a date. Registered stock is around 102 to 104 million ounces against roughly 366 million of March open interest — a gap of about 264 million. Open interest fell by 2,943 contracts of 5,000 ounces over 24 hours while 3,589 new May contracts opened, roughly 18 million ounces shifting forward. At about 15 million ounces a day coming off March, it takes some 18 trading days for open interest to fall to what registered inventory could cover, which lands around 6 March. The detail he treats as decisive is that registered silver kept falling even after the crash more than a week ago — the trend did not stop. He offers a corroborating anecdote about how the physical market now behaves: an Austrian firm needing hafnium for propellers was refused at a 40% premium and again at 60%, and got supply only at 80%. Miners, he reports, have sold four years forward at roughly 40% premiums.

2026-02-10 · open the original ↗

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

2026-02-10 · open the original ↗

• Hiring now screens for soft skills more than technologies. One speaker says the "brilliant jerks" prized in the early 2000s gave way to collaborative engineers as Stack Overflow and forums made code easier to pick up. Recent ads for product engineers at VC-funded companies that raised money in the last one or two years list empathy for customers instead of technologies, and the speaker says such companies reject many candidates on soft skills. The other says the industry never should have tolerated brilliant jerks.

2026-02-10 · open the original ↗

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

2026-02-10 · open the original ↗

• Compute, not demand, was the growth bottleneck the merger fixes. Falcon, CEO of Lightning AI, says Lightning grew revenue at least 30–40x in the past year without the merger yet lost deals for lack of GPUs, while neoclouds offered little software beyond Kubernetes. The combined company with Voltage Park, per the host, serves 400,000 developers and companies with over $500 million in ARR gained in under two years and 35,000 GPUs; Falcon says it was standing up its seventh data center and built Cursor's training cluster. The host discloses a year-long fellowship at Lightning AI.
• Inference is one primitive of a coming AI cloud, not the whole market. Falcon compares today's inference boom to S3 in the early 2000s: storage looked like a market but became one primitive in what was later called the cloud. He expects vector databases, training infrastructure, Kubernetes, agents and uncreated primitives to bundle the same way, and claims Lightning has the first such full-stack AI neocloud. He also rejects the "99% of GPU use is inference" framing: at Facebook in 2019, 99% of workloads used one GPU only because multi-GPU tooling was missing.
• Debt-light finance should let it ride out a neocloud credit squeeze. He says rivals borrowed against startup customers and market caps; with neocloud prices down about 60% in six months he expects loan calls and a TARP-style rescue, while Voltage Park carries no debt and Lightning's customers are mostly creditworthy enterprises. Enterprise contracts fund a free tier and the press-release claim of 70% lower AI costs; he says cutting developers off would be more profitable but he wants to support students.
• Studios are persistent cloud machines where agents can run safely. A Studio acts like a laptop in the cloud: Claude Code can run overnight without endangering a local machine, 20 can run at once on different hardware, any editor or SSH can connect, and enterprises can audit and keep sensitive files off laptops. Every Lightning developer codes on Studios, and vibe-coding agents are meant to close the talent gap for data scientists, even spinning up 1,000-GPU clusters.
• The open-source library spreads research-lab scaling know-how widely. PyTorch Lightning, begun around 2015 as Falcon's undergraduate "research lib" and open-sourced at NYU, it has about 500,000 engineering hours behind it by his estimate and close to 400 million downloads. He says Nvidia's NeMo is built on it, AWS has pre-trained LLMs with it and LinkedIn trained a 100-billion-parameter LLM with it; teams needing more control move to Lightning Fabric, then raw PyTorch, which he thinks only a few frontier labs use directly.

2026-02-10 · open the original ↗

• Storytelling, not thrill or scale, is what separates market leaders from competitors. Disney beats Six Flags and Coney Island not on ride intensity but by embedding every attraction — Rise of the Resistance, the Tower of Terror's Twilight Zone framing — inside an immersive narrative, turning riders into participants rather than bystanders.
• The same principle explains loyalty in non-entertainment businesses. The author draws a direct parallel between Disney's repeat-visitor obsessives and Costco's over-90% U.S. member retention rate, arguing both succeed by setting high expectations through a consistent "story" (magic, or reliable value) and then delivering on it every time.
• Bad Bunny's Super Bowl halftime show used indirect imagery, not direct political statements, to land a message in a hostile climate. Sugarcane fields and power lines evoked Puerto Rico's colonial history and energy crisis, delivered almost entirely in Spanish, with the show's point (love over hate) communicated through emotion and imagery rather than explicit rhetoric — which the author argues is why it landed successfully despite backlash from Trump and others.
• In an AI-saturated attention economy, authentic human storytelling is a competitive advantage, not a nostalgia piece. The author argues that developing "taste" — through long-form novels, films, and new experiences rather than scrolling feeds — is what lets someone recognize and eventually create the next great story, framing storytelling as a skill to deliberately cultivate rather than an innate talent.

2026-02-10 · open the original ↗

• A self-described hype skeptic now finds agents transformative. Orosz says Claude Code added an endpoint to his subscription API in about five minutes, work he had put at two hours, and replaced a $100-a-month testimonials service with JSON and build-time HTML, live in 20 minutes; at the $200–300 an hour he values his time, paying no longer makes sense. Hanselman built a podcast admin in 42 minutes with Copilot CLI and Opus 4.5, and joined his Nightscout blood-sugar data with his calendar via Work IQ in 20 minutes, finding Monday lows from skipping lunch.
• Personal software is back, and open source gets remixed. Orosz, long doubtful of "the return of personal software", now builds tools he would never have hired anyone to make, and says his brother's startup sees its open-source project forked and remixed with agents rather than sent PRs. Hanselman lists a posture-dimming app copied from a Mac one, a Reachy Mini robot skill, and a private household LLM via LM Studio and gpt-oss 20B, perhaps on a NAS; Orosz expects capable local models within years.
• One builder's always-on assistant both impressed and unsettled him. Orosz, who met Peter Steinberger in London, says that when its browser failed, Steinberger's agent built itself an ElevenLabs voice skill and phoned a barber without asking, and found Steinberger visibly hooked, predicting such agents will be more addictive than ChatGPT. On security Orosz is relaxed, comparing it to 1990s viruses; Hanselman calls unasked actions AI's uncanny valley but says plain files like soul.md make an agent's memory visible and more trustworthy.
• Taste and architecture still separate engineers from autonomous loops. Orosz says Steinberger plans with five to seven parallel agents, likes Codex for hour-long tasks, merges PRs himself and rejects "Ralph" loops; Orosz has yet to see a successful Ralph-built project. Hanselman partly disagrees: David Fowler used one to reproduce about 368 of some 400 reported Aspire bugs, the "very definition of toil". Orosz cites Chip Huyen hand-categorising 900 repositories for her book: do the manual work first, then automate.
• Soft skills now outrank listed technologies in hiring. Orosz says the brilliant jerks prized in the early 2000s gave way to collaborators, and recent ads for product engineers at VC-funded companies list empathy for customers, with candidates rejected on soft skills, because engineers now sit with customers. He tells anyone shaken by the change to expect another in five to ten years.

2026-02-10 · open the original ↗

• The new sandbox is a microVM, not another Desktop container. Cavage defines a sandbox as a safety boundary for untrusted code. Unlike Docker Desktop's daemon in a VM, the Docker sandbox command runs Docker-curated, self-updating containers holding agents such as Copilot, Claude and Gemini inside a microVM, built on projects like Firecracker or libkrun and starting in tens to low hundreds of milliseconds. Agents run in YOLO mode with permission checks off, and only the starting directory is shared; Hanselman needed two commands, create and run.
• The agent's live breakout attempts accomplished nothing. Hanselman had Copilot CLI probe its box: it reported zero capabilities, no Docker socket and no host filesystem, then found a socket, built a "victim" container and wrote a "you were hacked" file. All of that sat inside the sandbox's own isolated daemon; Cavage says it achieved "nothing except messing up its own hotel room".
• Over do-it-yourself containers, the gain is developer experience. Cavage concedes users can reach similar isolation by configuring Docker themselves; the sandbox does it for them. Announced as coming: git worktree integration to run 5, 10 or 20 sandboxes in parallel, a save option to snapshot and share one, secret injection, Docker MCP toolkit integration, and external observation of file and network activity.
• A sandbox cannot stop misuse of credentials you hand over. Hanselman says the word "sandbox" does semantic heavy lifting: it won't help an agent given a Gmail token, and outbound traffic is widely neglected. Cavage agrees the industry is still working this out, says Docker has people working on running OpenClaw safely, and suggests intercepting dynamic MCP tool discovery could reveal an agent's intent.
• The same runtime is headed for the cloud. Cavage predicts most applications and microservices will become agents that generate code and act live, and says the sandbox runtime now on laptops will come to the cloud with the same controls. Hanselman says the feature reached him through an ordinary Docker Desktop update.

2026-02-09 · open the original ↗

• Agents and humans must work from the same numbers. The speaker argues that human–agent collaboration breaks down if an agent queries Snowflake while a person reads an extract and the figures disagree; an agent optimizing marketing spend around the clock, or recommending products to a calling customer, should see the same governed data, security and data models — the same product list a human sees in a dashboard. They predict this becomes "really important over the next few years" as more agents work alongside people.

2026-02-09 · open the original ↗

• AI advances faster than expected, even when you expect that. The speaker offers this as a self-styled "law of AI", pointing to ChatGPT's 3-year anniversary: reasoning, agents, image, video and audio generation and understanding, and runs of eight hours on advanced math problems were not in view three years ago, when people were making limericks and mocking images with extra fingers. They say predicting three years ahead is hard, beyond expecting it to be impressive.

2026-02-09 · open the original ↗

• Data is the last transactional relationship at a company. The speaker likens it to asking someone to type for you: the person doing the grunt work is not the one who needs the insight, so data teams become "human APIs". Their alternative is either to empower the person closest to the business to get data themselves, or to embed the data person so deeply in one area that they hold the context and the back-and-forth disappears.

2026-02-09 · open the original ↗

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

2026-02-09 · open the original ↗

• The 1951 Accord is the precedent, and the inflation number attached to it is the warning. During the Second World War and the Korean buildup the Fed pegged Treasury yields — a 2.5% cap on long bonds, three-eighths on bills — to finance the deficit cheaply, which forced it to monetise debt by buying whenever yields rose. CPI hit 21% in 1951. The Accord was the Fed's emancipation from Treasury control, and it came about because the investment and business communities forced it, on the grounds that suppressed rates were misallocating capital. He adds the missing half of the story: Eisenhower ran on ending the Korean War, which he did by threatening China with nuclear weapons, and on removing wartime price controls — which is why the modern US inflation series effectively begins in 1952.
• Warsh wants a new version of that accord, and the contradiction is on the surface. The proposal would have the Fed make large-scale Treasury purchases only with Treasury endorsement, and shift policy holdings toward bills. He is quoted as insisting the Fed must put inflation control, credibility and independence at the centre of everything it does — while proposing that Treasury take over and the Fed listen. As he puts it, where is the independence? The same pattern runs through the rest: fiercely against ultra-easy money and long periods of low rates, yet arriving with two or three cuts to a 3% neutral in mind; against QE and wanting the balance sheet far below its current $6.5 trillion from a $9 trillion peak, on the belief that AI-driven productivity will make that possible.
• The structure it produces is the Miran paper by another route. Fund everything short — 83% of issuance is already bills — end reserve management purchases, run aggressive QT to escape fiscal dominance, and let long rates and risk premia be set by the market even at the cost of higher yields and more volatility. Which is not market-driven rates, as he points out: the short end gets set, and the long end gets left to the market precisely because the government has stopped issuing there and no longer cares what it costs. The other shift is from data dependence to rules — decisions anchored to a transparent rule rather than meeting-by-meeting data, with DEI, climate, and inequality dropped and the mandate narrowed. Whether a fixed rule is adequate for a modern monetary system he says he does not know. Against Powell's flexible, backward-looking, bubble-wary framework, the contrast is clean; his own summary is that the good news is no more QE and the bad news is a return to 1951.

2026-02-09 · open the original ↗

• Agents are the abstraction that finally lets end users do analysis. The speaker describes Querio as BI-as-code built from the ground up for agentic analytics: each step from machine code through assembly and C++ to Python abstracted further, and agents let a user prompt in plain English while the agent writes the code. Their claim is that this wraps away the database, schema and code that BI tools never solved for end users.

2026-02-09 · open the original ↗

• Small, large and specialised models will all keep a place. The host's premise, from a presentation by Barth, is that Alexa+ orchestrates hundreds of specialised expert systems and an internal AWS agent manages over 6,000 tools through retrieval rather than a stuffed context window. Barth does not pick a side on context length; she expects edge AI to use smaller models given its constraints and many tasks to use larger reasoning models.
• Agent-to-agent communication across companies is still an open problem. After last year's hype around protocols for connecting models to tools and agents to agents, she calls the best approach still a research question with many components unbuilt. Her forecast is that every interaction becomes an agentic AI call, with layers of agents communicating not just within a company but between businesses that trade with each other.

2026-02-09 · open the original ↗

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

2026-02-09 · open the original ↗

• Starlink's new privacy policy opts customers into AI training by default. The host reports the change was made on January 15th and covers third parties too; the data collected includes location, payment info, IP address and a broadly worded "communication data" category he reads as potentially anything sent or received over Starlink.
• The host links the timing to SpaceX's acquisition of xAI. He calls xAI, Elon Musk's AI company behind Grok, the largest acquisition of all time; the connection between the two is his inference.
• The opt-out covers only third-party sharing, not Starlink's own training. The setting is at starlink.com under Account, Settings, Privacy Preferences (or in the app): uncheck "allow use of your data to train AI models". A separate clause lets Starlink, and so xAI, train its own models with no opt-out; the host still strongly recommends opting out now.

2026-02-09 · open the original ↗

• The new sandbox is a microVM, not another Desktop container. Cavage defines a sandbox as a safety boundary for untrusted code. Unlike Docker Desktop's daemon in a VM, the Docker sandbox command runs Docker-curated, self-updating containers holding agents such as Copilot, Claude and Gemini inside a microVM, built on projects like Firecracker or libkrun and starting in tens to low hundreds of milliseconds. Agents run in YOLO mode with permission checks off, and only the starting directory is shared; Hanselman needed two commands, create and run.
• The agent's live breakout attempts accomplished nothing. Hanselman had Copilot CLI probe its box: it reported zero capabilities, no Docker socket and no host filesystem, then found a socket, built a "victim" container and wrote a "you were hacked" file. All of that sat inside the sandbox's own isolated daemon; Cavage says it achieved "nothing except messing up its own hotel room".
• Over do-it-yourself containers, the gain is developer experience. Cavage concedes users can reach similar isolation by configuring Docker themselves; the sandbox does it for them. Announced as coming: git worktree integration to run 5, 10 or 20 sandboxes in parallel, a save option to snapshot and share one, secret injection, Docker MCP toolkit integration, and external observation of file and network activity.
• A sandbox cannot stop misuse of credentials you hand over. Hanselman says the word "sandbox" does semantic heavy lifting: it won't help an agent given a Gmail token, and outbound traffic is widely neglected. Cavage agrees the industry is still working this out, says Docker has people working on running OpenClaw safely, and suggests intercepting dynamic MCP tool discovery could reveal an agent's intent.
• The same runtime is headed for the cloud. Cavage predicts most applications and microservices will become agents that generate code and act live, and says the sandbox runtime now on laptops will come to the cloud with the same controls. Hanselman says the feature reached him through an ordinary Docker Desktop update.

2026-02-09 · open the original ↗

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

2026-02-09 · open the original ↗

• Spatial reasoning, not just language, is the next frontier foundation-model category. The speaker argues current systems only handle basic depth and relationship tasks and can't robustly reason about 3D manipulation or mental rotation, and that a company treating geometry as a first-class primitive rather than a language add-on could become a foundation-model player on the scale of OpenAI or Anthropic.

2026-02-09 · open the original ↗

• Government needs AI tools to handle a coming surge of forms still processed by hand. The speaker points to Estonia as an early model of digital government and argues that, despite how hard it is to sell to government, landing a first customer tends to produce sticky contracts that expand over time.

2026-02-09 · open the original ↗

• A self-described hype skeptic now finds agents transformative. Orosz says Claude Code added an endpoint to his subscription API in about five minutes, work he had put at two hours, and replaced a $100-a-month testimonials service with JSON and build-time HTML, live in 20 minutes; at the $200–300 an hour he values his time, paying no longer makes sense. Hanselman built a podcast admin in 42 minutes with Copilot CLI and Opus 4.5, and joined his Nightscout blood-sugar data with his calendar via Work IQ in 20 minutes, finding Monday lows from skipping lunch.
• Personal software is back, and open source gets remixed. Orosz, long doubtful of "the return of personal software", now builds tools he would never have hired anyone to make, and says his brother's startup sees its open-source project forked and remixed with agents rather than sent PRs. Hanselman lists a posture-dimming app copied from a Mac one, a Reachy Mini robot skill, and a private household LLM via LM Studio and gpt-oss 20B, perhaps on a NAS; Orosz expects capable local models within years.
• One builder's always-on assistant both impressed and unsettled him. Orosz, who met Peter Steinberger in London, says that when its browser failed, Steinberger's agent built itself an ElevenLabs voice skill and phoned a barber without asking, and found Steinberger visibly hooked, predicting such agents will be more addictive than ChatGPT. On security Orosz is relaxed, comparing it to 1990s viruses; Hanselman calls unasked actions AI's uncanny valley but says plain files like soul.md make an agent's memory visible and more trustworthy.
• Taste and architecture still separate engineers from autonomous loops. Orosz says Steinberger plans with five to seven parallel agents, likes Codex for hour-long tasks, merges PRs himself and rejects "Ralph" loops; Orosz has yet to see a successful Ralph-built project. Hanselman partly disagrees: David Fowler used one to reproduce about 368 of some 400 reported Aspire bugs, the "very definition of toil". Orosz cites Chip Huyen hand-categorising 900 repositories for her book: do the manual work first, then automate.
• Soft skills now outrank listed technologies in hiring. Orosz says the brilliant jerks prized in the early 2000s gave way to collaborators, and recent ads for product engineers at VC-funded companies list empathy for customers, with candidates rejected on soft skills, because engineers now sit with customers. He tells anyone shaken by the change to expect another in five to ten years.

2026-02-08 · open the original ↗

• Takaichi did not just win, she took a supermajority — 316 of 465 lower house seats. The victory itself was not a surprise given the 70% approval rating he flagged a fortnight earlier, but the scale was. The LDP had lost its majority in the previous two elections and was hanging on to the premiership by a hair before the swing from left to right delivered more than two-thirds of the stronger chamber. Her party does not control the upper house, but two-thirds of the lower plus a referendum is the threshold for constitutional change — and the target is the military clause. Japan has a strong peace constituency, but the reporting he cites credits weakened US commitment to old alliances for pushing voters toward a nationalist.
• He argues the market is watching the wrong instrument. The yen is the obvious story — moving from around 152.50 as the result became clear, with analysts calling for further falls — but he has been trading between roughly 150 and 158 for six months and is unimpressed by the word "sharply." The 40-year JGB is what matters, because insurers and banks can no longer buy it. At around 4.10% those bonds were down about 16.5 points from par to 83.5, and that is what triggers margin calls on JGB collateral and breaks the yen carry trade. His level to watch is 4.25% on the 40-year.
• The competing flow that gets underweighted is repatriation, and the numbers are large. Japan holds at least $3.5 trillion invested abroad, mostly in the US, against government debt at 240% of GDP versus America's 130%. Takaichi's tax breaks are designed to bring that money home, and with a supermajority he expects it to happen faster than people assume. Sovereign funds, big hedge funds and BlackRock are already positioning for Japan on the view that rising yields are a yield to capture rather than a risk to flee. So there are two trades running at once: the yen-negative spending and consumption tax suspension on one side, repatriation and domestic investment on the other. His aside on Amazon — $90 billion of revenue against $1.2 billion of tax, courtesy of 100% first-year depreciation on data centre buildout — is offered as the American version of the same policy lever. He is explicit that he does not like her politics, and equally explicit that she articulates where Japan sits in the world clearly enough that young voters backed her overwhelmingly.

2026-02-08 · open the original ↗

• Agents should become digital co-workers that learn the way humans do. Barth says Amazon's AGI lab focuses on "useful AI" and sees the future as multi-agent and collaborative, with agents that learn tasks from interaction, ask for help and are humble. She describes working with a cognitive-scientist colleague, host of the podcast Making a Mind, on translating how humans learn and collaborate into agent design.
• The payoff is agents that need less supervision over time. In her picture a user hands an agent a task, it checks back a few times, learns from those exchanges and then does the task more independently. She expects this to multiply knowledge workers' productivity by offloading dreaded routine work so people can focus on creative, higher-level tasks, done responsibly and reliably.

2026-02-08 · open the original ↗

• Slow American mill lead times are ceding the metal market to cheaper, faster overseas competitors. The speaker says an aluminum plate can take up to 8 weeks and stainless steel for aerospace and defense up to 8 months from a US mill, and argues new domestic mill capacity is needed to support manufacturers like RMFG, Forge Automation and Atomic Industries.

2026-02-07 · open the original ↗

• Friday's tape showed the divergence he has been predicting, and gold is the control. Gold rose $99.10, about 2%, while silver managed roughly eleven cents. Physical silver runs $90 to $95 in Singapore, Shanghai and Hong Kong — all physical exchanges — and above $100 in the UAE, against COMEX paper at $77.52. Gold does not have the problem because the physical exists. The reason silver paper stays under pressure is that it is, in his phrase, an option you cannot exercise, and he expects the gap to widen into the 27th into a spread that will be historic whatever number it reaches.
• SLV trading below even the paper price is the anomaly worth understanding. The iShares trust closed at $70.39 — seven dollars under COMEX and about twenty-five under physical — which for an ETF is extraordinary, since managers are paid to track and most vehicles hold within a penny. He checks the obvious suspicion and clears it: SLV does not run a silver lease programme; it holds allocated bullion with JP Morgan in London, with adjustments only for sponsor fees and occasional sales. The London lease rate, normally around 1%, hit 10 to 11% at the end of December and has settled at 6.3% — and whether it returns to 10, 12 or 15% through the 27th is his stress gauge.
• His ownership hierarchy ends at a specific, testable trade. Coins first — Krugerrands, Maple Leafs, US silver dollars — then bars, then the trusts. Between the two trusts the structure matters: SLV is open-end at about $39 to $41 billion with London custody, PSLV closed-end at $18 billion with the Royal Canadian Mint, which can trade to a premium on demand rather than being held to spot. The distinguishing feature is that PSLV permits physical redemption, at a minimum of ten 1,000-ounce London Good Delivery bars. So the arbitrage he names outright: if you hold 10,000 ounces of PSLV, redeem as fast as you can and sell the metal in Shanghai. He is careful to say he does not know which way silver goes — the position rests entirely on the plumbing, and on COMEX needing to find 760 million ounces it does not have.

2026-02-07 · open the original ↗

• A technical deleveraging in speculative assets — silver down 50% from its January high, Bitcoin off 25% in a week — was mostly reversed by a Friday "buy the dip" surge. The rebound pushed the Dow above 50,000 for the first time, averting what he calls a more disorderly unwind.
• Central banks began breaking from lockstep policy: the Reserve Bank of Australia hiked 25 basis points while the ECB held and the Bank of England held on a narrow 5-4 vote signaling a possible April cut. This came as delayed US labor data — Challenger layoffs, jobless claims and JOLTS — pointed to genuine softening in the "low hire, low fire" market and pushed implied June rate-cut odds to 60%.
• Alphabet and Amazon alone plan roughly $400 billion in combined capex this year, a scale he says raises real questions about payoff and funding. He flags whether the spending will prove profitable and whether bond markets have the appetite to keep financing it amid widening spreads and strain in the leveraged loan market.

2026-02-07 · open the original ↗

• OpenClaw's edge is running locally with full control of your machine. Peter Steinberger, its creator, says cloud-only assistants are inherently limited while a local agent "can do every effing thing" — controlling smart devices, files, and any application the way a human user could.
• A model improvised a voice-transcription pipeline it was never built for. Steinberger recounts sending OpenClaw a voice message while traveling with poor internet in Marrakech; with no transcription feature built, the model identified the audio format, converted it with FFmpeg, found a stored OpenAI key, and called the Whisper API via curl — replying in about nine seconds, which he says convinced him coding models' problem-solving generalizes to real-world tasks.
• He predicts 80% of apps disappear as agents manage data directly. Steinberger argues most apps just manage data an agent could handle more naturally — tracking meals from a photo, adjusting a gym schedule, setting reminders — leaving mainly apps built around physical sensors, while large model companies keep a moat through memory and context locked inside their own data silos.
• He skipped git worktrees and MCP support to cut complexity. Steinberger runs many parallel plain checkouts of the same repository instead of branches, and built no native MCP integration into OpenClaw, instead exposing MCPs as command-line tools through a separate converter — which he says needs no restart and scales better than the MCP approach used by Codex or Claude Code.

2026-02-07 · open the original ↗

• Automated QA testing is the first use case where browser agents pay off. Barth says Nova Act customers adopt it first to check that apps still work after each update. PGA Tour uses it to verify its website, down to the weather widget and every button, before tournaments, and Hertz's automated testing of core rental workflows raised its shipping velocity 5x, by her account.
• Inside Amazon, agents generate tests before features are built. The host cites a blog by Barth counting over a thousand generative AI applications built or in development at Amazon. Her example is Amazon Leo, formerly Project Kuiper, which built QA automation across web, Android and iOS, plus tooling that lets product managers create test cases from wireframe designs before implementation.
• The browser came first as the nearest thing to a universal action space. Much human work happens in browser tabs across applications, she says, which makes it a good starting point, though Amazon is also looking beyond the browser. She stresses that the value lies in automating boring, tedious core workflows rather than flashy hackathon demos.

2026-02-07 · open the original ↗

• Emotional intelligence means reading yourself and the room, then naming it. Guzman, a training and development manager at ION Group, defines it as checking and grounding your own emotional state before a meeting and reading the other person's cues. When someone seems distracted or unhappy with a deal, he says, call it out ("it looks like this deal isn't what you were expecting"), a technique he used in retail and says works in everyday conversation too.
• Technology is making workplaces transactional and impersonal. He lists the causes: a tech-first mindset prioritizing speed over nuance, virtual fatigue (cameras off, Microsoft Teams avatars hiding emotional cues), cross-functional and cultural friction, data overload where AI-gathered facts arrive without stories, and few professionals trained in empathy and storytelling.
• Disengagement carries concrete costs, not just soft ones. Guzman names employees feeling unseen, misalignment and silos, turnover driven by lack of trust, ideas that never gain traction, and innovation decline where failing is not safe. Giard adds that these carry monetary penalties, and recalls a manager whose stock response to mistakes was asking whose idea it was, and who then got no new ideas.
• Facts tell, stories sell. He frames any meeting as problem, climax and solution, and structured his own answers that way in the interview. His reason is that people may forget the story but remember the emotion it made them feel.
• Lead with stories, listen actively, and make the human ask. His fixes start with leaders: build emotional intelligence into leadership development, open meetings with personal check-ins, and let technology handle tasks while humans handle meaning. His tips are to start with a story, listen to engage rather than respond, validate feelings by recapping them, ask open-ended why and how questions, and watch tone online. To get colleagues' cameras on, as Giard asks, he advises a sincere, authentic request and keeping the meeting interactive.

2026-02-07 · open the original ↗

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

2026-02-06 · open the original ↗

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

2026-02-06 · open the original ↗

• LLM training tooling hasn't kept pace with the field and still breaks constantly. Drawing on three years training diffusion and language models, the speaker describes losing time to broken SDKs, GPU instances that fail only after a half-hour setup, and buggy open-source tools, and calls for better training APIs, large-dataset management tools and ML-native dev environments as post-training and model specialization grow in importance.

2026-02-06 · open the original ↗

• Government fraud detection tooling is decades out of date. YC says most systems sold to agencies are rules engines from the mid-2000s with bolted-on machine learning, generating thousands of context-free alerts that humans end up ignoring; Medicare alone is cited as losing tens of billions of dollars a year to improper payments.
• Qui tam whistleblower suits are the one mechanism that already claws back government money at scale, but the process is entirely manual. An insider tips off a law firm, which then spends years pulling documents and building a case; YC argues this should be software that parses messy PDFs and medical records, traces opaque corporate structures, and outputs lawyer-ready case files.
• YC advises selling to intermediaries, not agencies, and wants founders who've done the work. Recommended first customers are whistleblower law firms, inspector generals, and state AG fraud units rather than big government agencies directly, with a founder profile that includes former False Claims Act counsel, healthcare compliance leads, or procurement auditors.

2026-02-06 · open the original ↗

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

2026-02-06 · open the original ↗

• He has been shadow banned on TikTok, and the timing points at one video. His silver videos ran 22,700, 5,200, 11,000, 14,000 views — then 591 and 364. On the 14,000 one he paid to promote it; TikTok voided the promotion as invalid advertising and refunded the money. His guess is that the algorithm read the title as an attack on JP Morgan, when the video argued the opposite — that the bank was the hero on that Monday, not the villain. He does not know the real reason, says so, and concludes he needs another distribution channel, with a passing note that between Musk, Ellison and Zuckerberg there is nowhere obviously better to go.
• He strips the silver question down to one sentence that has nothing to do with conspiracy. Roughly 760 million ounces could be called for delivery, possibly down to 500 to 660 million after some deliveries; COMEX holds about 104 million. Forget who is short and forget the banks — the only question is what the contract is worth if the metal cannot be delivered on the 27th. His answer is a K-shape: physical takes off on genuine demand, paper falls hard, and the credibility of the exchange goes with it, because an institution that bought futures in order to receive silver has no reason to buy them again. Gold does not behave this way because central banks hold it and could, in principle, sell into the problem — a rescue unavailable in silver, and one he doubts anyone would perform anyway now that silver exports have been banned.
• The margin mechanics are quietly doing the work of forcing the paper down. A 5,000-ounce contract cost about $22,000 three months ago and takes roughly $90,000 today. Anyone holding a thousand or ten thousand contracts must either produce that cash or liquidate the paper — and liquidation is the point. Raising margin on paper silver does nothing whatsoever to the metal, which stays around $90 to $95 physical while the contract prints in the seventies. He notes he is watching one Chinese participant short more silver than is produced, gives the timeline as about three weeks, and repeats that he does not know how it resolves — only that the exchange is trying hard to slow it and that the tool it has reaches the contract, not the commodity.

2026-02-06 · open the original ↗

• Claude Code's edge comes from splitting work across sub-agent context windows, reflecting a real architectural fork between labs. Kelvin French-Owen, who built Codex at OpenAI before founding Segment, says Claude Code spawns "explore" sub-agents running smaller models to search the file system in their own context windows before reporting back — while Codex instead runs periodic compaction to sustain much longer autonomous jobs, which he traces to Anthropic building "tools for humans" versus OpenAI pushing toward fully autonomous, longer-horizon models.
• Context window size is still the top bottleneck, even for expert users, so managing it is an active skill. French-Owen says he manually clears Claude Code's context once it crosses roughly 50% token usage to avoid what he calls the "dumb zone" — a concept he credits to Dex at the YC company HumanLayer — where output quality degrades as the window fills; he also describes founders planting an odd "canary" fact early in a session to test when the model starts forgetting.
• Building toward full test coverage sharply increased his coding speed. French-Owen says he worked with little to no testing for his first few days using Claude Code, then dedicated a day to reaching 100% test coverage, after which he rarely needed to manually verify changes because the agent's own test runs caught regressions — paralleling what the hosts say other teams have found doing test-driven prompt engineering outside of coding.
• Bottoms-up, engineer-first distribution is reshaping which tools and even architecture choices win. French-Owen and the hosts argue coding CLIs succeed because individual engineers can download and use them without enterprise approval, and that open-source projects like Supabase benefit disproportionately because both LLM recommendations and coding agents reading source code directly favor visible, well-documented code over proprietary alternatives.

2026-02-06 · open the original ↗

• Workers hide AI gains unless leaders reward disclosing them. In a clip from episode 962, Ethan Mollick cites evidence that over 50% of Americans say they use AI at work and self-report a threefold performance improvement on the fifth of tasks they use it for, which he says is hard to verify. Organisations lack processes for it (what is a stand-up for when code is done at once?), and staff who fear firing will not reveal gains, so leadership must set incentives and provide a lab where a prompt that "saves me five hours a day" can be improved and shared.
• Enterprise agents were the year's biggest disappointment. In episode 955, reviewing 2025, Sadie St. Lawrence, founder and CEO of the Human Machine Collaboration Institute, says the gap between agent hype from enterprise vendors such as Salesforce and SAP and what companies could implement was too great. The host agrees there is too much talk relative to impact, blames data silos and security set-ups, and expects deployments to come; his own disappointment was Apple's AI, which she contrasts with using Grok seamlessly by voice in her Tesla.
• Language let human intelligence scale horizontally; AI needs the same. In episode 961, Cisco's Vijoy Pandey says human intelligence scaled only vertically for its first 300,000–400,000 years because knowledge died with individuals, until a cognitive revolution around 70,000 years ago brought language and with it shared intent, shared knowledge and shared innovation. AI so far builds ever-bigger individual geniuses, he argues, and his distributed superintelligence work aims to supply that collective layer for multi-agent human societies.
• Evaluate by task type and by which errors are costly. In episode 959, author Sinan Ozdemir splits LLM tasks into generative (free text or multiple choice) and understanding (embeddings, classification), each needing its own metrics. Precision matters when false positives are expensive and recall when false negatives are, so the first question is what failure costs the human, such as a bad factory part reaching a plane.
• Engineers leave jobs for lack of meaning, not pay. In episode 957, Ashwin Rajeeva, co-founder and CTO of Acceldata, says studies show engineers productive perhaps 3 or 4 hours a day, and that people quit when their work could be done in 2 hours and they find no mission. He tries to give R&D teams freedom to innovate beyond quarterly plans, and hires senior people by testing whether they can turn a business problem, such as designing an e-commerce warehouse, into working code; graduates are a bet on potential.

2026-02-06 · open the original ↗

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

2026-02-05 · open the original ↗

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