May 2026
Perplexity chief business officer Dmitry Shevelenko argues that AI agents should be judged less as software features than as metered digital labor: tools users will pay for when they perform economically useful work. In a Big Technology Podcast interview, he makes the case that Perplexity’s computer-use agents, workflow packaging, broad permissions and multi-model orchestration are all part of that shift. The unresolved question is whether users and companies will accept the access, trust and usage-based pricing required to make those agents a real business rather than another AI novelty cycle.
Across today’s sources, applied AI was framed less as a contest over standalone models and more as an operating problem: agents need source, memory, monitoring, constraints, and secure access to do useful work. The same systems view appeared in infrastructure, where demand is spreading beyond GPUs into CPUs, memory, fiber, fabs, power, chip design, and platform control points.
Anthropic researcher Subhash Kantamneni presents Natural Language Autoencoders as a way to translate Claude’s internal activations — the numerical states produced while it answers — into readable text. The central claim is that this can expose what a model appears to be representing before it speaks, including whether a successful safety-test result reflects the intended behavior or recognition of the test itself. In Anthropic’s simulated blackmail evaluation, Claude refused to act harmfully, but the NLA translation suggested it also understood the scenario was likely a safety evaluation.
OpenAI is presenting three new API audio models as infrastructure for voice applications that can translate, transcribe, reason and act in real time. Romain Huet’s demonstration centered on GPT-Realtime-Translate, which keeps pace with multilingual speech, and GPT-Realtime-2, a voice-agent model that can follow turn-taking instructions, use business context and call tools while explaining its work. GPT-Realtime-Whisper completes the set as a streaming speech-to-text model for live transcription.
Michael Arnaldi, of Effectful, argues that coding agents use Effect better when the project gives them the Effect source code, not just better prompts or documentation. In a workshop starting from an empty repository, he demonstrates cloning the Effect repo into the project, having the agent extract local pattern files, and then using strict TypeScript diagnostics, tests, lint rules and persistent instructions to steer the agent toward a working Effect HTTP API.
Arm chief executive Rene Haas told Bloomberg Tech that weakening smartphone demand is being offset by a faster-growing AI data center business, where order visibility for Arm’s AGI CPU has doubled to $2 billion in five weeks. Haas argued that agentic AI workloads are increasing the need for CPUs to handle orchestration and scheduling that GPUs cannot manage, making Arm’s opportunity less dependent on handset volumes and more tied to data center infrastructure, supply-chain execution and rack-level power efficiency.
Arman Kolozyan of the Max Planck Institute for Security and Privacy argues that many zero-knowledge proof bugs arise from a mismatch between what a prover’s computation can produce and what a verifier’s constraints will accept. Presenting the paper “Language-Agnostic Detection of Bugs in Zero-Knowledge Proof Programs,” he describes a formal model, the Domain Consistency Model, and a static-analysis tool, CCC-Check, designed to detect those mismatches across ZKP languages. The work reports two-orders-of-magnitude speedups over SMT-based approaches on benchmarks and 15 previously unknown vulnerabilities in six ZKP projects, most of them outside the reach of existing models.
Hoover Senior Fellow John Taylor recounts the Mont Pelerin Society’s 1947 founding as an organized response to the postwar rise of socialist planning and state intervention. In Taylor’s telling, Friedrich Hayek and the society’s first members were not simply defending markets in economic terms; they were trying to rebuild classical liberalism as an intellectual program for freedom, aimed at the practical policy questions that had made collectivism newly credible.
Chris Williamson’s debate with Freya India, William Costello and Tania Reynolds uses polling on young women’s negative views of men as the starting point for an evolutionary-psychology account of modern dating. The panel argues that older sex differences have not disappeared but have been pushed into online politics, group chats, beauty markets and relationship norms, where mate choice, vulnerability, in-group loyalty and public moral signaling make relations between young men and women more adversarial.
Tech humanist Stephen Remedios built “DaddyGPT,” an AI version of himself, to handle his three sons’ routine permission requests while he worked. The problem began when it worked: his children kept using the bot even when their parents were beside them, because it was always available, calm and adaptive. Remedios argues that AI’s risk in parenting and other care relationships is not only failure, but convenience that displaces the imperfect human presence those relationships require.
Raindrop’s workshop argues that production agents need a different observability model from conventional software monitoring or offline evals. Zubin Kumar, Danny Gollapalli and Ben Hylak make the case that teams should track both explicit telemetry such as tool errors, latency and cost, and implicit signals such as user frustration, refusals, task failure, capability gaps and unusual workarounds. Their framework treats real production behavior as the primary surface for finding regressions, running experiments and catching failures that do not appear as clean exceptions.
Dick Costolo’s account of running Twitter from 2010 casts the company’s early crisis less as founder drama than as an operating failure: too many decisions required group consent, too few people clearly owned outcomes, and process was allowed to substitute for judgment. In a conversation with Brian Halligan, the former Twitter CEO argues that scaling the company required replacing consensus with accountable decision rights, a bias to yes, direct communication, and faster correction when mistakes or personnel problems became clear.
Replit founder Amjad Masad told Sam Parr and Shaan Puri that Replit’s jump from roughly $2.5 million to $250 million in revenue run-rate was not a smooth growth curve but the result of a market-creation moment. In his account, Replit Agent turned years of stalled platform ambition into a product non-engineers could use to build, deploy and run software, producing about $1 million of ARR on its first day and changing the company’s problem from finding demand to keeping up with it.
AJ Andrews argues on Bloomberg’s The Deal that baseball and softball can grow by giving fans more access to athletes’ personalities, backstories and identities, not by adding artificial spectacle. Andrews points to the World Baseball Classic as proof that visible pride and emotion can deepen fan attachment, while Alex Rodriguez makes a parallel case that baseball leaves value unused by hiding players’ preparation and personal stories. For softball, Andrews says the lesson is exposure: when games and players are put on television and treated as distinct, compelling stories, audiences have a reason to follow.
Consumer investor and marketer Rohan Oza argues that breakout food and beverage brands are built by upgrading large existing habits, not inventing new ones. In a My First Million interview with Sam Parr and Shaan Puri, Oza lays out the playbook behind Vitaminwater, Smartwater and Poppi: create desire through culturally credible influencers, translate that demand into retail shelf space, keep the economics intact, and negotiate hard when a strategic buyer such as Coke or Pepsi comes calling.
Bloomberg Technology framed AMD’s sharp rally as evidence that the AI infrastructure trade is widening beyond GPUs. Caroline Hyde, Ian King and RBC’s Srini Pajjuri said AMD’s forecast pointed to renewed demand for CPUs as AI workloads shift toward inference and agentic systems, even as Nvidia remains dominant in accelerators. The program extended that argument across Nvidia’s Corning deal, Microsoft’s power constraints and Apple’s outside-model plans: the AI boom is becoming a contest over compute, connectivity, energy and platform control.
Bloomberg’s primer argues that the AI boom is turning the semiconductor supply chain into a strategic stress test, raising demand for advanced processors while exposing how dependent the industry remains on a handful of companies, machines and manufacturing clusters. The source traces that pressure through ASML’s lithography tools, AMD’s AI chip designs, TSMC’s concentration of advanced fabrication in Taiwan, and competing US and Chinese efforts to rebuild domestic capacity. Its central claim is that chips are becoming more economically and politically essential just as their production remains physically fragile, capital-intensive and difficult to replicate.
Economists Steven Davis and Evan Starr examine whether states should enforce noncompete agreements, a contract tool Starr says now reaches well beyond executives and trade-secret holders to low-wage workers, interns, and janitors. Starr argues the evidence points to weak worker consent, continued use of unenforceable clauses, and lower innovation where noncompetes are enforced; Davis presses the countercase that some clauses may protect legitimate firm investments in training, clients, or confidential information. Their disagreement centers on whether law can distinguish those uses cheaply enough, or whether broader bans and bright-line rules are the better response.
Elizabeth Economy’s conversation with Randy Schriver and Mike Kuiken of the US-China Economic and Security Review Commission argues that Washington’s China problem now cuts across trade, technology, supply chains, cyber operations, Taiwan planning, pharmaceuticals, and sanctions policy. Schriver and Kuiken say the US government still manages many of those risks through agencies and laws built for an earlier era, leaving economic statecraft fragmented just as China’s leverage has become more integrated. Their case is less for severing all economic ties than for building the machinery to decide which ties are tolerable, which are dangerous, and which require national effort to replace.
Mark Gurman said Apple has held early talks with Intel and Samsung about using new U.S. fabs to make future A-series and M-series processors, an exploratory move he framed as a supply-chain redundancy question rather than only a political one. Apple still relies heavily on TSMC, primarily in Taiwan, and Gurman described that geographic and supplier concentration as one of the company’s biggest risks. Across the rest of the broadcast, executives and analysts described a similar shift from exposure to execution: AI companies are giving Washington early model access for review, while enterprise adoption is being tested by security, deployment cost and proprietary data advantages.
Aspen FSP’s recording argues that the affordability crisis makes wealth building more urgent, not less, because families cannot achieve durable stability on constrained cash flow alone. Joanna Smith-Ramani and Genevieve Melford frame stability and wealth as mutually reinforcing, while leaders from Compass Working Capital, GRO and Stackwell point to programs that combine income support, externally funded capital, trusted guidance and routes into appreciating assets. Their case is that low-wealth households need the same kinds of balance-sheet supports that higher-income families often receive through employer benefits, tax advantages and private wealth.
Aspen Digital’s B Cavello argues that AI evaluations should be understood by philanthropy as a way to shape the AI ecosystem, not merely as technical measurements or benchmark leaderboards. In a briefing for philanthropic leaders convened with Siegel Family Endowment, Cavello says funders can influence what AI developers optimize for, support outside accountability through audits and related tools, and help users judge when systems are appropriate for their needs.
Bloomberg’s Sangmi Cha argues Samsung’s move past a $1tn market value is more than a symbolic milestone: traders are reading it as a direct expression of the AI infrastructure trade, driven by tight memory-chip supply and helped by news of an Apple partnership. Cha says the rally still has room in investors’ eyes because Samsung trades at about 5.3 times forward earnings, while the company’s surge is also feeding a broader foreign-led rally in Korean equities.
Eyck Freymann’s argument in Defending Taiwan, discussed with Niall Ferguson at the Hoover Institution, is that U.S. deterrence is too narrowly built around stopping a Chinese invasion. Freymann says Beijing could instead use customs controls, coast guard pressure, energy constraints, supply-chain leverage, and political coercion to force Taiwan toward submission without triggering a clear war. His prescription is for Washington to build credible options between inaction, military escalation, and an economic rupture it cannot sustain.