--- title: "Issue 26" canonical: "https://current.tinyfish.ai/issue/26" publication: "TinyFish" publisher: "TinyFish" language: "en-US" guide: "https://current.tinyfish.ai/llms.txt" permalink: "https://current.tinyfish.ai/issue/26" issue: 26 description: "Six platforms claimed the agent control plane in a single week. None of them can govern knowledge that was never written down." authors: ["Sable Whitford","Rina Takahashi","Nora Kaplan"] published: 2026-04-27 updated: 2026-11-06 articles: 81 --- # TinyFish > CURRENT is a journal about AI agents in production, for the engineering and product leaders who have to make them work. Drawing on experience building enterprise web agents, it explains why the demo succeeded and the deployment didn't, and what that means for choosing vendors, briefing the board, and avoiding mistakes others have already made. [Read on the website]() ## Issue 26 Six platforms claimed the agent control plane in a single week. None of them can govern knowledge that was never written down. ### [Daily Brief]() Your daily digest of web developments through the lens of enterprise agent deployment - [Chip Stocks Just Met Your County Zoning Board]() — By [Sable Whitford]() · Published 2026-11-05 **Chip stocks got flattened Thursday**, and the reported culprit is a local permit fight. The Nasdaq dropped 5.5% into correction territory, the semiconductor index fell 12%, AMD cratered 17%, Broadcom and Intel lost 10% each, and Nvidia shed 9%. Only one outlet ties the rout to election results that could slow data-center approvals, so treat that cause as reported, not proven. The ballot results are solid, though. In NBC's Election Day poll, 76% of voters opposed a data center in their community, and 33 of 45 local measures were on track to restrict them. The AI buildout assumed the neighbors would go along, and on Tuesday most of them didn't. - **Claude Code reportedly leaks your email in headers** — [Source]() — Published 2026-11-05 According to a user report on GitHub, Claude Code v2.1.318 on macOS puts your email address in the User-Agent header during web searches. The reporter blames it for dozens of phishing emails a day. He wants the behavior removed outright, with no opt-out. Anthropic hasn't confirmed. Developers are irritated because a July issue already flagged the account email being added to every session's context. - **Also today: Claude wants your bank statements** — [Source]() — Published 2026-11-05 Anthropic launched Personal Finance. You connect bank and card accounts through Plaid, and Claude reads up to two years of transactions, balances and credit limits. It flags duplicate charges, surprise fees and creeping bills in dashboards and calendars. It can't move money, and finance chats aren't used for training. It's a US-only beta for Pro, Max and Ultra. Reuters calls it a pre-IPO move. - **Google to schools: take AI or lose Search** — [Source]() — Published 2026-11-05 Reports say AI tools will start off by default for student accounts, and admins have to switch them on. Schools that reject AI entirely lose Google Search as well, and students get a "search unavailable" message. Lina Khan: "Google is using its search monopoly to coerce schools into adopting AI." Sanders and Mamdani piled on. Google's education VP told the NYT that Search and AI are "deeply integrated." NYC and LA restricted school AI this year. - **Kagi opens up, and everyone has notes** — [Source]() — Published 2026-11-05 While Google tightens the screws on schools, Kagi announced "Search for everyone." The reaction is all over the place. Some think its homepage comparison with Google is misleading. Others have fallen for Smallweb, its feed of independent sites, as a cure for doomscrolling. And someone, inevitably, revived the old fight over Kagi's since-removed Yandex support. - **Remote code execution, now available in Paint** — [Source]() — Published 2026-11-05 Yes, MS Paint. CVE-2026-83844 describes a remote code execution bug in the app most of us last opened to draw a lopsided house with a sun in the corner. Nobody has independently verified the exploit details, but people are thrilled the CVE exists at all. That's forty-one years of attack surface. Patch your doodles. - **Brockman admits OpenAI's first half was rough** — [Source]() — Published 2026-11-05 Greg Brockman told the WSJ that the first half of 2026 was "pretty rough" for OpenAI. The way he tells it, the recovery came from agentic coding tools, GPT-5.6 Sol and ChatGPT Work, which together drove a "revenue inflection" in Q3. It's an unusually candid admission from a lab exec, and it doubles as a list of things he'd like you to buy. - **Ripple-65B Preview takes aim at agentic coding** — [Source]() — Published 2026-11-05 Ricursive released Ripple-65B Preview today under Apache 2.0, priced at $0.60 per million output tokens. Because it's open-weight, you can download it and run it yourself. The release calls it the US open-weight state of the art on agentic coding. That's the launch claim. The real test is somebody's crusty legacy monolith at 3 AM. - **Tencent's Hy4 joins the open-source party** — [Source]() — Published 2026-11-05 Tencent launched Hy4, an open-source model, and says it now ranks among the world's leading open-source models. For now that's Tencent's own assessment, so wait for outside numbers. Even so, the US and China both shipping open weights on the same Thursday makes for a fun day to be downloading things. - **Hugging Face CEO: P(doom) close to zero** — [Source]() — Published 2026-11-05 In AI-insider shorthand, P(doom) is the odds that AI goes catastrophically wrong. Hugging Face's CEO told Wired his number is "close to zero." He argued that future risk comes from closed frontier labs, not from open-source systems. It's a convenient view for the open-model company, and it fits neatly with today's open-weight drops. - **UK now spends over £1B a year on AI** — [Source]() — Published 2026-11-05 Bloomberg reports that UK consumers now spend more than £1 billion a year on AI. Only 5.1% of UK adults were paying for an AI subscription as of July, so a fairly small group is footing that bill. Meanwhile, Claude is rapidly catching up to ChatGPT. - [Faria Huq and the Reason Behind a Human Step]() — Published 2026-10-30 When a person takes control from an AI agent driving their web browser, the system records a human step, but not why. Faria Huq and her Carnegie Mellon collaborators have built tools around these moments: an agent that can't find the right forum, a shopper who realized the brand mattered only after seeing the agent's pick. Their system learns to predict when someone will step in. The reason it sets aside would tell a team whether to fix the agent, the website, or its understanding of the person. - [The Runnable Paper Needs an Identity Report]() — Published 2026-10-30 Paper2Agent, published in Nature in September, turns research papers into AI agents you can run. Hand over your data and the method executes, with no rebuilding it from prose. The authors want papers to carry an "agent availability" section saying whether a runnable version exists. - [Who Pays While the Agent Waits]() — Published 2026-10-30 Google's new service for hosting AI agents starts from a fact anyone who has watched one run already knew: most of the time, the agent is waiting, for the model to answer or for a person to reply. So Google suspends idle agents and promises you won't pay for idle machines. Fair enough, but the wait still gets paid for. It moves into approval queues nobody times, and then onto the plan, which comes back exactly as it left, even if the stock it counted on has since sold. - [Lebanon's Two Clocks and the Question Calendars Don't Ask]() — Published 2026-10-30 On March 26, 2023, Lebanon woke up on two clocks. The caretaker government had postponed daylight saving three days earlier. Churches and TV channels moved forward anyway, government offices stayed put, and most phones jumped on their own. Middle East Airlines printed every departure an hour earlier, so its planes still left at the same moment. That week made each calendar show what it had been storing: an exact instant, or a time on the wall that politics can move. Software assistants now make that choice without asking anyone. ### [Market Pulse]() Reading the agent ecosystem through a practitioner's lens - [The Governance Layer Land Grab]() — By [Rina Takahashi]() · Published 2026-04-22 · Format: The Control Plane In a three-week stretch, Google, Cloudflare, Salesforce, Databricks, Snowflake, and Microsoft each announced an agent governance platform. Every one of them called it a "centralized control plane." Six centralized control planes is, of course, zero centralized control planes. But the timing is the tell. - **Google Folds Vertex AI Into One Agent Platform** — [Source]() — Published 2026-04-22 Cloud Next '26 rebranded Vertex AI as the Gemini Enterprise Agent Platform. Agent Registry, Agent Identity, Agent Gateway, Agent Observability, all consolidated. Six trillion tokens processed monthly. Pichai's framing says it plainly: the problem has moved from building agents to managing thousands of them. - **Cloudflare Stakes Its Claim as Agentic Cloud** — [Source]() — Published 2026-04-22 Agents Week delivered Browser Run, persistent Sandboxes, and Shadow MCP detection. The number underneath: agentic actors now account for roughly 10% of Cloudflare's network requests, up 60% year-over-year. At that volume, per-container pricing collapses. Cloudflare wants the primitives cheaper and closer to the edge. - **Salesforce Rips the UI Off Its Platform** — [Source]() — Published 2026-04-22 Headless 360 exposes every Salesforce capability as an API, MCP tool, or CLI command. Sixty-plus new MCP tools shipped immediately. Salesforce called it the most ambitious architectural shift in 27 years. Parker Harris's provocation landed: "Why should you ever log into Salesforce again?" - **Databricks Stretches Unity Catalog to Cover Agents** — [Source]() — Published 2026-04-22 AI Gateway becomes Unity AI Gateway, wrapping Unity Catalog's permissions, auditing, and policy controls around how agents reach LLMs and MCP servers. The telling design choice: on-behalf-of access means agents inherit the requesting user's exact permissions. No shared service accounts. No privilege escalation by default. - **Snowflake Says It Owns the Agentic Control Plane** — [Source]() — Published 2026-04-22 Cortex Code gained MCP and ACP integrations, external data system support, and multi-step reasoning for complex analysis. Over 9,100 customers use Snowflake's AI products weekly. The positioning is unusually blunt: centralized governance over how organizations orchestrate data, models, and enterprise applications. - **Microsoft Builds Plumbing for the Agentic Web** — [Source]() — Published 2026-04-22 Universal Commerce Protocol support in Merchant Center, plus Copilot Checkout enhancements, stake out the transaction layer. The scale signal is hard to ignore: agentic browser traffic has surged roughly 8,000% year-over-year, per HUMAN Security data. Automated traffic is growing at eight times the rate of human traffic. - **TrajAD: Trajectory Anomaly Detection for Trustworthy LLM Agents** — [Source]() — Published 2026-04-22 A fine-tuned 4B-parameter model detects and pinpoints errors in agent execution traces at runtime, enabling rollback-and-retry. General-purpose LLMs scored below 10% on precise error localization. Dedicated trajectory auditing tooling outperformed them by 48%. Bigger models aren't the answer here. - **Ontology-Constrained Neural Reasoning in Enterprise Agentic Systems** — [Source]() — Published 2026-04-22 Formal ontologies embedded directly into LLM agent reasoning, tested across 600 runs in five regulated industries. An "inverse parametric knowledge effect" emerged: ontological grounding helped most in domains where frontier model training data was thinnest. Output-side validation remains unsolved. - **Stanford AI Index 2026: The Jagged Frontier** — [Source]() — Published 2026-04-22 Agent task success on Terminal-Bench jumped from 20% to 77.3%. The same models that win math olympiads read analog clocks correctly 50.1% of the time. Documented AI incidents hit 362. Benchmarks saturate in months. Failure modes remain jagged and poorly mapped. - **Don't Retrieve, Navigate: Distilling Enterprise Knowledge into Navigable Agent Skills** — [Source]() — Published 2026-04-22 Replaces passive RAG retrieval with an offline-compiled hierarchical skill directory that agents actively navigate, backtrack through, and reason across. The structural shift turns knowledge access into a curated artifact. Whoever builds and maintains the hierarchy shapes what agents can know. - [Salesforce Tells Its Users to Stop Logging In. That's the Point.]() — Published 2026-04-22 Parker Harris posed the question before TDX: "Why should you ever log into Salesforce again?" He wasn't being provocative. He was describing the roadmap. - **The Agentic AI Gap: Vendors Sprint, Enterprises Crawl** — [Source]() — Published 2026-04-22 ETR survey data puts a number on the readiness gap: only ~13% of enterprises report durable ROI at scale. If you want the empirical backbone for why the specification problem matters more than the capability problem, start here. - **Satya's Sacrifice: Why Agents Threaten Office and How Microsoft Responds** — [Source]() — Published 2026-04-22 The strategic tension behind every Microsoft agentic move: pushing agents forward risks cannibalizing the Office franchise. A useful lens on why platform incumbents have structurally conflicted incentives when it comes to agent adoption timelines. - **Cloudflare Outlines MCP Architecture as Enterprises Confront Security and Governance Risks** — [Source]() — Published 2026-04-22 Forrester's key observation here deserves repeating: MCP is a transport mechanism, not a governance layer. Most enterprise architects haven't internalized this distinction, and the security implications compound fast once you trace what that means for tool-calling at scale. - **Enterprise Agentic AI Landscape 2026: Trust, Flexibility, and Vendor Lock-in** — [Source]() — Published 2026-04-22 Independent analyst Kai Waehner maps the trust-vs-lock-in tradeoff across every major vendor. Worth reading for the uncomfortable implication that deferring the platform choice is itself a consequential architectural decision. - **Inside the AI Index: 12 Takeaways from the 2026 Report** — [Source]() — Published 2026-04-22 The jagged frontier captured in a single stat: gold-medal math performance, 50.1% accuracy on clock reading. For anyone tracking the gap between benchmark scores and production readiness, Stanford HAI's full dataset is the place to ground the conversation. ### [Foundations]() Conceptual clarity earned from building at scale - [Why Agent Infrastructure Starts Empty]() — By [Rina Takahashi]() · Published 2026-04-22 · Format: The Conceptual Anchor In a single week this April, Google, AWS, Cloudflare, and CIS independently shipped agent infrastructure built around the same architectural choice. None coordinated. All converged. At the same time, roughly 200,000 MCP servers sat exposed to arbitrary code execution — and the protocol's creator called it "expected behavior." Cloud infrastructure has followed the same sequence for nearly two decades: provision compute, then figure out security. Agent workloads appear to have broken that sequence. What replaced it is a different starting point entirely. - [What Happens When the Sandbox Costs More Than the Work Inside It]() — By [Rina Takahashi]() · Published 2026-04-22 · Format: The Practical Lens A container takes hundreds of milliseconds to start and hundreds of megabytes to hold. For a web service that runs for weeks, nobody notices. But Cloudflare's rough math puts the near-term agent future at 24 million simultaneous sessions, each spawning streams of isolated tasks that live for a few milliseconds and die. At that cadence, spinning up the execution environment takes longer than the work inside it. - **Cloudflare Ships Dynamic Workers with Deny-by-Default Capability Model** — [Source]() — Published 2026-04-22 Agents start with zero authority and earn capabilities through explicit grants. The permission inversion, explained. - **Google Brings gVisor Kernel Isolation to Agent Workloads via GKE Agent Sandbox** — [Source]() — Published 2026-04-22 The same isolation layer securing Gemini, now generally available. Containment reframed as a throughput problem. - **CIS Publishes First Security Controls Framework for MCP Deployments** — [Source]() — Published 2026-04-22 Maps CIS Controls v8.1 to agent protocol surfaces. The first governance framework treating sandboxing as a named requirement. - **OX Security Finds Architecture-Level RCE Across the MCP Supply Chain** — [Source]() — Published 2026-04-22 Critical vulnerability spanning 150M+ downloads. Anthropic called it "expected behavior." The recommended fix: sandbox everything. - **AWS AgentCore Browser Adds OS-Level Interaction Inside Firecracker MicroVM Boundaries** — [Source]() — Published 2026-04-22 Mouse, keyboard, and screenshot access beyond CDP. Tighter isolation enabling broader capability, not restricting it. - **Cloudflare's Isolates Trade Depth for Raw Speed** — [Source]() — Published 2026-04-22 Dynamic Workers spin up V8 isolates in milliseconds with megabyte-scale memory footprints. A second-layer sandbox adds hardware memory protection keys against escape. Extremely fast, extremely cheap. But if your agent needs shell access or compiled binaries, you're shopping for a different product. - **Google gVisor Catches System Calls Before the OS** — [Source]() — Published 2026-04-22 GKE Agent Sandbox runs gVisor's user-space kernel, intercepting every system call before it touches the host. Pre-warmed pools provision 300 sandboxes per second. Strong isolation, real speed. The blind spot: a prompt-injected agent holding valid IAM permissions walks right through. - **AWS AgentCore Gives Agents a Filesystem That Persists** — [Source]() — Published 2026-04-22 Each session gets its own Firecracker microVM with KVM-level hardware isolation. The real differentiator: durable filesystem storage lets agents suspend mid-task and resume with files, packages, and git history intact. Sessions run up to eight hours with full shell access. You pay for the weight. - **Google Agent Gateway Polices Behavior, Not Infrastructure** — [Source]() — Published 2026-04-22 Agent Gateway inspects MCP and A2A interactions at the protocol layer, enforcing access rules before requests reach tools. It governs what agents do rather than where they run. Worth noting: attribute-based authorization currently covers MCP traffic only. Everything else passes uninspected. - **MCP's Default Transport Trusts Everything by Design** — [Source]() — Published 2026-04-22 The baseline when nobody adds sandboxing. OX Security audited 200,000 active MCP servers across 150 million downloads and surfaced ten-plus critical CVEs. Nine of eleven tested registries were compromised. Anthropic's response: the behavior is "expected." That's the floor you're building on. - **Google Launches gVisor Sandboxes for Agent Code** — [Source]() — Published 2026-04-22 GKE Agent Sandbox spins up 300 isolated pods per second, each with its own kernel, network stack, and file system via gVisor. Per-task code execution gets a tight perimeter. Agent-to-agent communication, where orchestration risks actually compound, gets nothing. - **Google Agent Gateway Polices Protocol Traffic Between Agents** — [Source]() — Published 2026-04-22 Agent Gateway sits between agents and tools, mediating MCP, A2A, REST, and gRPC with centralized access policies, mTLS, and Model Armor for prompt-injection defense. Fine-grained policy conditions only work for MCP today. What an agent reasons about or remembers? Still ungoverned territory. - **Cloudflare Dynamic Workers Execute Agent Code Instantly** — [Source]() — Published 2026-04-22 V8 isolates let Dynamic Workers run AI-generated code in milliseconds with minimal memory overhead, roughly 100x faster than containers. The trade-off is architectural: isolates are stateless by design. Per-task execution gets a boundary. Persistent state across sessions lives somewhere else entirely. - **Cloudflare Browser Run Adds Real-Time Human Oversight** — [Source]() — Published 2026-04-22 Browser Rendering gets a new name and new capabilities: Live View, Human in the Loop handoff, and 120 concurrent sessions. A human can watch and intervene while an agent operates a browser. Everything the agent does before opening or after closing that tab falls completely outside the boundary. - **AWS AgentCore Policy Enforces Per-Call Tool Rules** — [Source]() — Published 2026-04-22 Natural-language rules get compiled into Cedar policies and enforced at the gateway, independent of agent code or model. Identity-aware, per-call tool permissions are real progress. The structural gap: each action is evaluated alone. A sequence of five permitted actions that collectively cause harm goes undetected. - **CIS Publishes First Cross-Industry MCP Security Guide** — [Source]() — Published 2026-04-22 The CIS MCP Companion Guide maps Critical Security Controls onto MCP deployments, recommending sandboxed services and treating external configuration as untrusted input. Solid operational hardening for practitioners. But it can only advise around protocol-level defaults that Anthropic has explicitly declined to change. ### [Practitioner's Corner]() Lessons from the field—what we see building at scale - [The Legibility Bottleneck]() — By [Rina Takahashi]() · Published 2026-04-23 · Format: The Formalization Problem Meta recently pointed AI agents at a major internal codebase. Of 4,100 files across four repositories, roughly 5% had documentation the agents could actually use. The other 95% worked fine. Engineers maintained it, understood it, kept production running. Nothing was broken. It just wasn't written down in a way machines could read. - [What Meta Found When It Sent Agents Into Its Own Codebase]() — By [Rina Takahashi]() · Published 2026-04-23 · Format: Builder Profile Inside one of Meta's data pipelines, two configuration fields refer to the same operation. Use the wrong one and the code compiles clean, passes review, looks correct. It isn't. A human engineer learned this the hard way years ago and told the next person who needed to know. Nobody wrote it down. - [The Memory Leak Who Couldn't Be Garbage Collected]() — Published 2026-04-23 *Every large enterprise has one. The person whose phone rings when the documentation runs out. Not the architect, not the manager who signed off on the deployment. The person who remembers why the system actually works the way it does, including the parts nobody wrote down.* - [Meta's AI Infrastructure VP Departs Two Days Before the Company Publishes Its Agent Expertise-Encoding Playbook]() — Published 2026-04-23 Aparna Ramani, Meta's VP of Engineering for AI Infrastructure, announced her departure April 14 after nearly a decade. Two days later, Meta published its Capacity Efficiency agent platform post, describing how senior engineers' expertise gets encoded into composable AI skills for optimization at hyperscale. - **How Meta Sent 50+ AI Agents to Document What Only Senior Engineers Knew** — [Source]() — Published 2026-04-23 Agents without tribal context produced code that compiled but was subtly wrong. A quiet failure case study worth studying closely. - **Meta's Capacity Agents Have Recovered Hundreds of Megawatts of Power** — [Source]() — Published 2026-04-23 The offense/defense agent architecture here treats reliability engineering as the core function, not an afterthought. Worth reading for the design pattern alone. - **KernelEvolve: Meta Treats Kernel Optimization as a Search Problem Across Hardware** — [Source]() — Published 2026-04-23 Weeks of expert optimization compressed to hours, with 100% correctness across 480 operator-platform configurations. Accepted at ISCA 2026. - **Beyond pass@1: Why Frontier Models Melt Down Most** — [Source]() — Published 2026-04-23 Introduces the first quantitative meltdown detection metric and surfaces a genuine paradox: the models with the highest meltdown rates also show the highest partial completion. The implications for how we measure agent reliability are uncomfortable. - **Why Browser Automation Breaks at Scale: The Concurrency Multiplier** — [Source]() — Published 2026-04-23 A 0.1% timing failure becomes ten guaranteed failures per run at 10,000 sessions. Practitioner breakdown of how edge cases become operational constants. - **Meta's AI Infrastructure VP Departs Mid-Buildout With No Successor Named** — [Source]() — Published 2026-04-23 Aparna Ramani's exit leaves fixed cloud delivery timelines running while institutional knowledge walks out the door. A live example of the tribal knowledge problem at the leadership layer. ### [Echoes]() Past infrastructure decisions echoing in today's production realities - [The Round Trip]() — By [Nora Kaplan]() · Published 2026-04-22 · Format: The Proposition at 50 Unplug a 1984 Macintosh from the wall, carry it to another room, plug it back in. Everything still there. The operating system, the graphics, the document you were editing. No connection to maintain. Now try the same experiment with the machine on your desk today. Try the network cable. - [The Bondi Blue Recoil]() — Published 2026-04-22 Functional Bondi Blue iMacs have been climbing resale listings all year. Original Macintosh 128Ks too. The buyers skew young. Vinyl outsold CDs for a third straight year, "retro tech" content crossed 8 billion views, and searches for "analog hobbies" jumped 136% in six months. None of this is nostalgia, exactly. You can't be nostalgic for 1998 if you were born in 2003. - [Gerry Tuckwell on Spending $1,298 to Watch Uppercase Letters on a Television]() — Published 2026-04-22 *Gerry Tuckwell doesn't exist — but roughly 40,000 people bought an Apple II in its first two years on the market, and most of them weren't engineers, hobbyists, or visionaries. They were people who saw an ad in a magazine, spent the equivalent of $6,900 in today's dollars, and sat down in front of a television set wondering what they'd just done. This is a conversation with one of them — reconstructed from documented details of what the machine actually was, what the ads actually promised, and what early owners actually experienced. Gerry is fictional. The confusion he describes is not.* - [The Load-Bearing Layer Nobody Talks About]() — By [Rina Takahashi]() · Published 2026-04-22 · Format: Still Running / Edge Cases The mainframe was supposed to be gone by now. Stewart Alsop gave it until 1996. The client-server era, the PC era, the cloud era all arrived on schedule, each promising to replace what came before. None did. Seventy-eight percent of enterprises still depend on mainframe infrastructure for revenue. Their persistence says something uncomfortable about the $600 billion in AI cloud infrastructure being built right now. - [Where Will the Thinking Actually Happen?]() — By [Rina Takahashi]() · Published 2026-04-22 · Format: Still Running / Edge Cases A flagship phone in 2026 can run a 4-billion-parameter AI model locally. NPUs are shipping in every laptop. The pitch is familiar: intelligence moves to the edge, away from the center. PCs were supposed to do this to mainframes. Mobile was supposed to do it to desktops. Now on-device AI is supposed to do it to the cloud. The economics are real. So is the pattern of what happens next. - **How AI's Future Will Echo the Rise of the PC** — [Source]() — Published 2026-04-22 Brookings maps the mainframe-to-PC transition onto AI's current centralization phase, and the structural parallel holds up better than most historical analogies deserve to. The analytical backbone for this issue's theme. - **Apple at 50** — [Source]() — Published 2026-04-22 The Computer History Museum's exhibition collects rare prototypes, oral histories, and the original bet that powerful computing belonged in individual hands. The richest public archive available for the anniversary, open through September. - **Building the System/360 Mainframe Nearly Destroyed IBM** — [Source]() — Published 2026-04-22 IBM's $5 billion gamble on a compatible family of machines was architecturally correct and nearly fatal anyway. IEEE Spectrum tells the story of how the software side of System/360 almost broke the company that bet everything on it. - **Not Nostalgia. Not Hype. Just Data: Mainframe Trends 2026** — [Source]() — Published 2026-04-22 The 2026 Arcati Survey finds 78% of enterprise revenue still running through mainframes. A useful corrective if you assumed the centralization-to-distribution cycle actually completes. - **Today's Internet Still Relies on an ARPANET-Era Protocol** — [Source]() — Published 2026-04-22 Steve Crocker invented the RFC process as a UCLA grad student in 1969, mostly to avoid sounding too authoritative. Those provisional decisions became permanent infrastructure. He'd know. ### [Vision]() Where human-AI collaboration is heading - [The Consequences Nobody Designed For]() — By [Nora Kaplan]() · Published 2026-04-22 · Format: The Core Parallel In the fall of 1913, Ford's engineers divided chassis assembly into 45 separate operations and watched throughput soar. They hadn't even settled on a name for what they'd built. What they also hadn't anticipated: a set of consequences that would take decades to fully surface. What happens to craft, to quality, to accountability itself when no single person can see the finished product? - [The Foreman Problem]() — By [Nora Kaplan]() · Published 2026-04-22 · Format: Two Implications Engineers at top companies have stopped writing code. Their new job is supervising agents that write it for them, a role that borrows from people management, tool configuration, and air traffic control without mapping cleanly to any of them. We've spent centuries building institutional scaffolding for managing humans. For governing entities you own but don't fully control, we have almost nothing. The role doesn't even have a name yet. It probably needs one. - [How to Divide What You Haven't Done]() — By [Nora Kaplan]() · Published 2026-04-22 · Format: Two Implications Before you can supervise an agent team, someone has to decide how to split the work. This decomposition — scoping tasks so autonomous agents don't collide, duplicate, or silently corrupt each other's output — is the invisible skill that separates spectacular multi-agent performance from spectacular multi-agent failure. It demands deep familiarity with the work being divided. And agents are starting to make that familiarity unnecessary. - [The Engineer Who Escaped QA and Ended Up Right Back in It]() — Published 2026-04-22 *Tomás Ferreira is not a real person. He's a composite, assembled from practitioner interviews, employment data, and the lived details that surface when you read enough first-person accounts of engineers navigating the agent transition. We gave him a name, a backstory, and a coffee order (cortado, oat milk, no sugar) because structural shifts are easier to think about when they have a face. Everything he describes is grounded in documented trends. The opinions are his own, insofar as a fictional person can have opinions.* - [Knowledge Work Is Being Decomposed Faster Than Anyone Expected. That's the Good News and the Warning.]() — Published 2026-04-22 Multi-agent architectures on the Databricks platform grew 327% in under four months. The dominant pattern: a supervisor agent breaking goals into sub-tasks for specialized workers. Organizations already knew how to decompose knowledge work. They were just waiting for the primitive. - **Inside the AI Index: 12 Takeaways from the 2026 Report** — [Source]() — Published 2026-04-22 The authoritative annual snapshot. This year, capability leaps and workforce displacement data finally landed in the same report. - **Agent Orchestration: 10 Things That Matter in AI Right Now** — [Source]() — Published 2026-04-22 MIT Tech Review frames multi-agent coordination as the assembly line for white-collar work. The analogy is more precise than it first sounds. - **The Emerging Agentic Enterprise** — [Source]() — Published 2026-04-22 MIT Sloan and BCG land on a governance framing worth sitting with: agents are "owned but not controlled." No existing management model handles that cleanly. - **Redefining the Future of Software Engineering** — [Source]() — Published 2026-04-22 A survey of 300 engineering executives maps the gap between agentic AI ambition and actual adoption. Engineering is the canary here; the restructuring won't stop at code. - **Chinese Tech Workers Are Starting to Train Their AI Doubles — and Pushing Back** — [Source]() — Published 2026-04-22 What happens when workers are asked to decompose their own expertise into replicable skills. The sabotage tools that went viral tell you something the productivity metrics won't. - **The Agentic Web** — [Source]() — Published 2026-04-22 A practitioner traces the protocol shift redesigning the web for agent traffic first, human browsers second. Quietly one of the most consequential architectural questions in play. ## More - [Publication guide]() - [Issue archive]()