--- title: "Issue 6" canonical: "https://current.tinyfish.ai/issue/6" publication: "TinyFish" publisher: "TinyFish" language: "en-US" guide: "https://current.tinyfish.ai/llms.txt" permalink: "https://current.tinyfish.ai/issue/6" issue: 6 description: "The web's architecture is forking again—this time between sites built for humans and infrastructure designed for agents. Every major vendor just picked their side of the split." authors: ["Sable Whitford","Rina Takahashi","Nora Kaplan"] published: 2025-12-08 updated: 2026-11-05 articles: 94 --- # 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 6 The web's architecture is forking again—this time between sites built for humans and infrastructure designed for agents. Every major vendor just picked their side of the split. ### [Daily Brief]() Your daily digest of web developments through the lens of enterprise agent deployment - [Google Lets Schools Keep Search Without Gemini. The AI Overviews Come Along Anyway.]() — By [Sable Whitford]() · Published 2026-11-04 On Tuesday, which was also Election Day, Google announced a K-12 policy that switches Gemini, Gemini Notebook and **Search off by default** for schools using its education software. A Google education VP told the New York Times that "Search and AI are deeply integrated," so districts that ban AI would lose Search as well. Bernie Sanders called it **"extortion,"** and Sen. Richard Blumenthal and NYC Mayor Zohran Mamdani joined in. Within hours Google **partly walked it back**: schools can now turn on Search by itself, without Gemini. That Search still puts AI Overviews on the results page, though, so **the fight now is over what "AI off" means.** - **Proxy Traffic Leaving the White House Network?** — [Source]() — Published 2026-11-04 Layer3 Intel says residential proxy traffic exited AS2650, the Executive Office of the President's network, 236 times in 30 days through 28 proxy services. It says it wasn't even looking. Its guess is that someone working late has a device carrying a proxy SDK. Treat that as a theory: no device has been named, and six public IPs can sit in front of a lot of machines. - **DreamHost Had Uninvited Guests Since August** — [Source]() — Published 2026-11-04 DreamHost confirmed that attackers used a file manager bug to steal session tokens, customer emails and files, and that they got in back in August. That's a long stay at a host with about 400,000 paying customers. DreamHost says it's unrelated to PHP2Shell. If you're a customer, rotating credentials today is a sensible move. - **PHP2Shell's Hangover Gets a Sequel** — [Source]() — Published 2026-11-04 PHP2Shell was a pair of 9.2-severity bugs that put NASA, Wikipedia and the White House at risk. Two weeks after the patches, Kaspersky counts nearly 30 million servers still unpatched, plus a third, still-open flaw (CVE-2026-115560). Working exploits keep multiplying, though there's no confirmed mass-breach wave yet. Tolmo's Matt Suiche says AI bug hunting is exposing 40 years of legacy code, and that's today's big argument. AI can find a bug in minutes. Someone still has to book maintenance windows for 30 million servers. - **Can DNS Tell You It's a Bot?** — [Source]() — Published 2026-11-04 Singapore's IMDA is reportedly exploring, with GoDaddy, Cloudflare and Google, whether DNS could identify AI agents. That would give the internet's 1983 phone book a second job checking IDs at the door. Every ops person already knows where the postmortem will land: it's always DNS. - **Meta Buys Conductor, Cloud Users Get Notice** — [Source]() — Published 2026-11-04 Meta bought Conductor, the app for running several coding agents in parallel, and the team is joining Meta Superintelligence Labs. Terms weren't disclosed. Conductor Cloud goes read-only Nov. 13 and shuts off Nov. 20. The local app gets model updates and critical fixes until early 2027. Conductor ran agents from rival labs too, so everyone wants to know whether that survives the move. - **OpenAI Shops Too: Pierre Computer** — [Source]() — Published 2026-11-04 Not to be outdone, OpenAI told Axios it plans to buy Pierre Computer, which stores and manages AI-generated code. It's OpenAI's second coding pickup after the Ona deal announced in June. Agents are producing mountains of code nobody typed by hand, and someone has to keep track of it. OpenAI would like to be that someone. - **Kimi K3.1 Pro Weights Drop: Bring Disks** — [Source]() — Published 2026-11-04 Moonshot launched Kimi K3.1 Pro in October, and today the open weights landed on Hugging Face. It has 2.8 trillion parameters with 104B active, and reportedly ships as 96 shards totaling about 1.56 TB under a custom non-OSI licence. Self-hosting starts with a purchase order for storage. Before the release, Artificial Analysis said it would cut the gap between open and proprietary models from 12 points to 3. - **Ripple-65B Crowns Itself America's Coding Champ** — [Source]() — Published 2026-11-04 Ricursive calls its Ripple-65B Preview "U.S. Open-Weight SOTA on Agentic Coding," meaning best in its class. That's the company grading its own homework. The "U.S." qualifier is carrying extra weight today, since a Chinese lab just released 1.5 terabytes of open weights. - **Go Gets a Brand-New JSON API** — [Source]() — Published 2026-11-04 The Go team introduced encoding/json v2, a new official JSON API, and Gophers have plenty to say about it. Nearly every Go service reads or writes JSON somewhere, so a second version of the standard package lands on the desk of anyone shipping APIs. - **Apple Puts Engineers on an AI Diet** — [Source]() — Published 2026-11-04 Bloomberg reports that in October, Apple cut some engineers' daily AI spend caps from $300 to $200. Some could drop to $100, and certain non-engineering teams could be capped at $25. Some teams are being nudged toward cheaper or on-device models. The punchline: few people ever hit $300, and Apple's R&D rose 32% in fiscal 2026. - **Texas Voted, and Data Centers Flinched** — [Source]() — Published 2026-11-04 Last brief, the data center backlash was winning local ballot fights. Now it has the Texas governor's mansion, in a state holding 25%+ of planned US capacity. Gina Hinojosa ran on an immediate moratorium, tougher water and power rules, and no tax breaks, and AI data center stocks fell on the result. GLJ Research says about 13 GW is exposed. Altimeter's Brad Gerstner says 10–15 GW might never get built. Abbott had already paused permits. Musk, who spent big on the losing side, now likes "data centers in space." - **Big Tech Pulls Out the AI Credit Card** — [Source]() — Published 2026-11-04 Oracle filed to sell $20B of new stock and bonds for its AI buildout, and Meta is seeking about $25B in bonds. The timing is bold. They're raising money to build on the same day the state with the most planned capacity elected a governor who ran on a moratorium. - **ChatGPT Search Is 20% of Google's?** — [Source]() — Published 2026-11-04 OpenAI's Adam Fry said Monday that ChatGPT search volume is about 20% of Google's, and that weekly users reportedly went from 1B to 1.2B in 60 days. Check the counting rules first. Any prompt that triggers a web search counts, and OpenAI didn't disclose the Google baseline or its method. Ahrefs puts ChatGPT at 0.36% of web referral traffic. - [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. - [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. - [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. ### [Market Pulse]() Reading the agent ecosystem through a practitioner's lens - [What Happens When Billion-Dollar Valuations Meet Production Reality]() — By [Rina Takahashi]() · Published 2025-12-03 · Format: The Signal Decagon raised $131 million in June at a $1.5 billion valuation—a 150x revenue multiple—on the assumption that AI agents will rapidly replace human support teams at massive scale. Five months later, Cognizant's chief AI officer stood at Web Summit and said something quietly devastating: "Their valuation is based on bigger is better, which is not necessarily the case." - [Enterprises Deploy Agents Faster Than They Can Measure Them]() — Published 2025-12-03 Agent deployments quadrupled from Q2 to Q3, hitting 42% of enterprises. Researchers discovered WebArena—used by OpenAI and others—marked "45 + 8 minutes" as correct. Top agents score 5% on hard benchmarks. Organizations are racing to production anyway. - **Enterprise Agents Fail Two-Thirds of Complex Tasks** — [Source]() — Published 2025-12-04 ServiceNow's AgentArch benchmark shows top models succeed on just 35.3% of complex enterprise tasks. More revealing: different foundation models demand different architectures. The one-size-fits-all agent design assumption breaks down under measurement, with immediate consequences for anyone building production systems. - **Machine Learning Discovers Better Agent Architectures Than Humans** — [Source]() — Published 2025-12-04 ICLR 2025 work introduces Meta Agent Search, automatically discovering architectures that beat manual designs across math, reasoning, and comprehension. Agent design follows the familiar ML pattern: human intuition gets replaced by learned solutions. The shift from engineering problem to optimization problem is underway. - **Agents Can't Tell When They're Wrong** — [Source]() — Published 2025-12-04 Systems-theory analysis reveals LLM agents suffer from metacognitive myopia: they struggle with environment interfaces, lack commonsense reasoning, and can't detect their own errors. The research argues holistic design matters more than component improvements. Emergent capabilities require architectural thinking, not just better prompts. - **Building Infrastructure for Agent Ecosystems** — [Source]() — Published 2025-12-04 Research maps infrastructure requirements for widespread deployment: inter-agent protocols, identity systems, certification mechanisms, action rollback. The focus shifts from individual agents to ecosystem connective tissue. What's needed isn't better agents but the infrastructure that lets them coordinate safely at scale. - **OpenTelemetry Standardizes Agent Observability** — [Source]() — Published 2025-12-03 The GenAI SIG finalized semantic conventions for agent telemetry, creating consistent metrics and trace attributes across frameworks. Standards work that enables production monitoring at scale. You won't appreciate this until you're debugging a multi-agent system at 3am. - **MCP's Connector Ecosystem Reaches Critical Mass** — [Source]() — Published 2025-12-03 Model Context Protocol's March updates added OAuth 2.1, JSON-RPC batching, and HTTP streaming. But the real story is over 1,000 open-source connectors, transforming a protocol specification into infrastructure enterprises can actually deploy. - **PagedAttention Halves GPU Memory Requirements** — [Source]() — Published 2025-12-03 vLLM's KV cache management with PagedAttention reduces GPU memory usage by 50% for long sequences, enabling larger batch sizes and higher throughput. Optimization work that makes production economics viable instead of aspirational. - **Prompt Compression Achieves 20x Token Reduction** — [Source]() — Published 2025-12-03 LLMLingua compresses prompts by 95% while preserving semantic meaning, reducing repetitive system prompts from 800 tokens to 40. Unglamorous optimization that compounds across millions of queries, turning cost centers into sustainable products. - **LLM Gateways Abstract Multi-Provider Complexity** — [Source]() — Published 2025-12-03 Unified gateways handle provider differences in APIs, authentication, rate limits, and pricing, enabling intelligent routing to optimal models per query. Middleware that solves the operational friction of managing OpenAI, Claude, Gemini simultaneously. - **ModelOps Frameworks Enable Enterprise Deployment** — [Source]() — Published 2025-12-03 ModelOps provides centralized lifecycle management, automated policy enforcement, and audit trails across ML, GenAI, and Agentic AI. Governance infrastructure that enables enterprises to deploy AI systems 10x faster with acceptable risk profiles. - **IBM: The AI Agent Boom Is FOMO-Driven** — [Source]() — Published 2025-12-03 Finally, someone admits most enterprises deploy agents because competitors did. The failure patterns here matter more than the hype. - **LangChain's State of AI Agents Report** — [Source]() — Published 2025-12-03 Half claim production deployment, but their write-in responses expose teams without basic testing. That gap tells you everything about market maturity. - **AWS Unveils Kiro: Autonomous Agent for Days** — [Source]() — Published 2025-12-03 Multi-day autonomy sounds impressive until you ask what breaks first. Context windows aren't the bottleneck when reliability architecture is missing. - **Microsoft: Making AI Agents Glass Boxes** — [Source]() — Published 2025-12-03 Observability determines whether you can run agents in production or just watch expensive demos fail. The architecture details here actually matter. - **Databricks: Stop Running Vibe Checks on Agents** — [Source]() — Published 2025-12-03 Walking through happy paths tells you nothing. Their systems approach to evaluation reveals what breaks before customers find it first. - **Production-Ready Agents: Design Principles That Work** — [Source]() — Published 2025-12-03 Autonomy without reliability just means unpredictable at scale. Architectural patterns here handle hundreds of tasks without constant human intervention. ### [Foundations]() Conceptual clarity earned from building at scale - [Browser Automation vs. Web Agents]() — By [Nora Kaplan]() · Published 2025-12-03 · Format: Field Guide A practitioner watches months of working Puppeteer scripts shatter overnight when a website redesigns. Selectors break. The workflow dies. Teams keep making the same mistake: treating browser automation and web agents as interchangeable tools for the same job. At scale, that confusion creates systems that fail in ways you didn't anticipate, or costs that compound when volume hits. The difference isn't just technical. It determines whether your automation survives production. - [When Frameworks Handle Timing for You]() — By [Rina Takahashi]() · Published 2025-12-03 · Format: Tools in Context Every web automation framework decides who handles timing ambiguity: the tool or the operator. Playwright chose automatic. Click a button and the framework waits for it to become interactive without manual coordination. Less control, simpler operations. When you're maintaining automation across dozens of sites with varying loading patterns, that choice determines whether you're fixing broken waits or building new capabilities. Abstraction overhead versus maintenance burden—which cost matters more depends on what you're trying to operate. - [When Explicit Control Reduces Overhead]() — By [Rina Takahashi]() · Published 2025-12-03 · Format: Tools in Context Puppeteer exchanges 11KB of WebSocket messages with Chrome to complete a task. Playwright needs 326KB for the same operation. That 30x difference comes from architectural choices: direct protocol access versus cross-browser compatibility layers. When your content loads predictably and Chrome is your only target, you're paying overhead for capabilities you don't need. Puppeteer gives you protocol-level control and native speed. For teams running high-volume automation against stable sites, explicit coordination beats automatic handling—if your production context rewards precision over convenience. - [Every Major Vendor Just Shipped Agent Orchestration]() — Published 2025-12-03 Between October and early December, Microsoft unified AutoGen and Semantic Kernel. OpenAI launched AgentKit with visual workflow builders. Google released ADK at Cloud NEXT. AWS announced AgentCore hit 2 million downloads in five months. Adopt AI shipped their framework too. - **What Breaks When Volume Hits?** — [Source]() — Published 2025-12-03 Your demo handles ten requests beautifully. Production throws ten thousand at it and suddenly you're debugging infinite loops at midnight. What specific failure modes emerge under real load? Not theoretical edge cases. The actual ways this breaks when you can't restart it every hour. - **Does This Become Infrastructure or Homework?** — [Source]() — Published 2025-12-03 Instagram served millions with 13 engineers because they chose boring PostgreSQL. Some technologies fade into the background. Others create new on-call rotations and splinter your team's expertise. Six months in, are you still babysitting this thing or has it become invisible? - **Does Staging Lie to You?** — [Source]() — Published 2025-12-03 Your staging environment passes every test. Then production reveals problems you never saw coming. Does staging actually replicate production's data volumes and traffic patterns, or are you testing in fantasy conditions? The gap between them determines what surprises you at launch. - **Can You Actually Monitor This Deploy?** — [Source]() — Published 2025-12-03 Someone needs to watch production for hours after deployment. Not because you expect disaster, but because you need to catch problems fast. Can you deploy when your team is actually available to respond? Friday afternoon deploys sound convenient until something breaks. - **Boring or Bleeding Edge?** — [Source]() — Published 2025-12-03 That shiny new framework promises everything. Proven technology has documented failure modes and available expertise when things break. What's the real cost of choosing novelty over boring reliability? The tradeoff shows up in operational predictability and how fast your team ships. - **Do Teams Share Definitions of Ready?** — [Source]() — Published 2025-12-03 Two-thirds of engineering leaders cite inconsistent standards as their biggest blocker. One team's "production ready" is another's "barely tested." Does your organization have shared definitions, or does every team interpret readiness differently? Inconsistency creates both risk and chaos. - [A Conversation with Semantic Sam, the Web's Quiet Revolutionary]() — Published 2025-12-03 We're sitting in what Semantic Sam insists on calling "a well-structured conversation space." Not a coffee shop, not an office, but somewhere with clear landmarks and proper hierarchy. Sam arrived exactly on time, which shouldn't surprise me. Punctuality is sort of their whole thing. ### [Practitioner's Corner]() Lessons from the field—what we see building at scale - [The Page That Lies]() — By [Rina Takahashi]() · Published 2025-12-03 · Format: Field Dispatch You click a button. Nothing happens. You click again. Still nothing. Then, three seconds later, both clicks register at once and you've accidentally submitted a form twice. The button was visible, styled, perfectly clickable. It just wasn't connected to anything yet. - [When Your Agent Fails at Step 47]() — By [Rina Takahashi]() · Published 2025-12-03 · Format: Builder Profiles When an agent fails at step 47 of a 50-step workflow—maybe verifying a hotel booking across a regional site that just changed its confirmation page—you don't get a stack trace. You get silence, or worse: a "success" message with subtly corrupted data that won't surface until a customer complains three days later. - [Browser Agents Hit 89% Success Rate, Humans 95.7%]() — Published 2025-12-03 The best browser agents now achieve 89% success on standardized benchmarks. Humans hit 95.7% on identical tasks. Six point seven percentage points. - **Config File Doubles, Internet Goes Dark** — [Source]() — Published 2025-12-03 November 18: Cloudflare's Bot Management config doubled in size, hit hard limits. X, ChatGPT, Shopify offline simultaneously. Engineers spent hours chasing Workers KV before finding the real problem. A configuration change took down a chunk of the internet. - **Protection Collapsed 67% This Year** — [Source]() — Published 2025-12-03 Testing 17,000 domains: full bot protection dropped from 8.4% to 2.8% in twelve months. Detection effectiveness ranged 6% to 42% depending on tool. Even large enterprises (10,001+ employees) left 61% of domains completely unprotected. The numbers are brutal. - **95% of Attacks Come From Authenticated Sessions** — [Source]() — Published 2025-12-03 Authentication stops nothing. 95% of API attacks in 2025 came from authenticated sessions. Injection and broken authorization made up over one-third of incidents. Organizations test only 38% of APIs, leaving legacy endpoints exposed. Stripe's /v1/sources breach proves the pattern. - **$44.5 Billion Wasted on Unused Resources** — [Source]() — Published 2025-12-03 Q2 2025 cloud spend hit $99 billion, up 25% year-over-year. But 21% goes to unused or under-utilized resources. That's $44.5 billion annually. One in three IT leaders report wasting over a quarter of their cloud budget. Scale amplifies waste. - **LLM Crawlers Quadrupled in Eight Months** — [Source]() — Published 2025-12-03 January to August 2025: LLM crawler traffic more than quadrupled. Some publishers saw web scraping reach eight times the volume of compliant AI bots when using advanced detection. The fundamental composition of automated traffic has shifted. Old assumptions about bot behavior no longer hold. - **Task Queue Chokes on Connection Surge** — [Source]() — Published 2025-12-03 Railway's November 25 incident: Temporal-backed task queue couldn't handle hundreds of thousands of new connections in minutes. All Free, Trial, and Hobby deployments stopped. Pro deployments continued with delays. The proxy layer plus metadata management hit scalability limits nobody anticipated. - **What Actually Breaks at Scale** — [Source]() — Published 2025-12-04 Real incident reports from Cloudflare, Google, CrowdStrike. Null pointers, validation bugs, cascading failures. The unglamorous reality of production. - **Google's Postmortem Culture** — [Source]() — Published 2025-12-04 At scale, incidents are inevitable. What separates mature operations from chaos is formalizing how you learn from failure. - **Running Postmortems That Actually Work** — [Source]() — Published 2025-12-04 Blameless is the easy part. Atlassian's guide covers the operational details: timing, facilitation, documentation that uncovers root causes. - **Browser-Use Framework Benchmarks** — [Source]() — Published 2025-12-04 Open source automation framework showing 3-5x speed improvements. Real code, concrete numbers. Worth examining if you're building at scale. - **Multi-Agent Coordination Beats Single-Agent Approaches** — [Source]() — Published 2025-12-04 93% on Mind2Web benchmark through specialized agent teams. The architectural insight: cardinality as a scaling dimension in web automation. - **Deploying Web Agents With LangChain** — [Source]() — Published 2025-12-04 Technical walkthrough addressing authentication, CAPTCHA, reliability. Shows the operational decisions you face when agents hit production constraints. ### [Echoes]() Past infrastructure decisions echoing in today's production realities - [When HTML Forked: The Compromise That Never Ended]() — By [Nora Kaplan]() · Published 2025-12-03 · Format: Infrastructure Inflections In February 1996, the HTML Editorial Review Board struck a deal: Netscape would remove blink tag from the specification if Microsoft removed marquee tag. Both companies agreed. Browsers kept supporting them anyway. - [WebSocket Broke Every Assumption About Connections]() — Published 2025-12-03 December 2011: RFC 6455 standardized WebSocket. Developers celebrated real-time communication without polling hacks. Infrastructure teams got a problem that still hasn't resolved. - **Complexity Always Gets Buried** — [Source]() — Published 2025-12-03 Kubernetes hid container orchestration. Platform engineering teams now hide Kubernetes. By 2026, 80% of large organizations will run internal developer platforms hiding the platforms that hide the infrastructure. Watch any technology long enough: someone builds an abstraction layer. Then someone abstracts that abstraction. The cycle never stops. - **The S-Curve Reveals Everything** — [Source]() — Published 2025-12-03 Slow start, explosive middle, grinding plateau. Every technology follows this arc. Early gains feel effortless. Later improvements demand exponential effort for marginal returns. AI adoption in 2025 sits somewhere on that curve. The shape never lies about where you are in the cycle or what comes next. - **Complexity Compounds Faster Than Solutions** — [Source]() — Published 2025-12-03 Add three nodes, get nine connections. Deploy microservices, coordination overhead explodes geometrically. The distributed monolith exists because coupling persists despite architectural intentions. Infrastructure drift becomes inevitable. Management overhead grows faster than the systems being managed. Always has, always will. - **Infrastructure Outpaces Human Learning** — [Source]() — Published 2025-12-03 Three point four million cybersecurity jobs sit unfilled globally. Hybrid clouds need expertise in both legacy and cloud systems simultaneously. Platform engineering emerged partly because skilled operators are scarce. Each technology wave amplifies the gap. Infrastructure evolves on Moore's Law. People don't. - **Centralize, Decentralize, Repeat** — [Source]() — Published 2025-12-03 Mainframes to PCs to cloud to edge. Gartner predicts 75% of enterprise data at the edge by 2025, up from 10% in 2018. Each swing solves yesterday's problems while creating tomorrow's. Edge computing is just the latest decentralization wave. The pendulum never stops moving. - **Performance First, Bills Later** — [Source]() — Published 2025-12-03 Build fast, discover costs afterward. Cloud spending hit $290 billion in 2023, growing 18% while organizations struggle to justify basic network upgrades. Infrastructure drifts toward poorly optimized states. The cycle repeats: optimize for performance, face cost pressure, re-architect, create new performance problems. Different decade, same pattern. - **CAP Theorem Proved the Impossible Triangle** — [Source]() — Published 2025-12-03 Brewer's 2000 conjecture, proven by Gilbert and Lynch in 2002, revealed what distributed systems builders already suspected: you cannot simultaneously guarantee consistency, availability, and partition tolerance. The theorem didn't create this constraint. It formalized an impossibility that was always there, forcing every database architect since to choose which guarantee to sacrifice. - **Fielding's Dissertation Documented the Web's Hidden Architecture** — [Source]() — Published 2025-12-03 Roy Fielding's 2000 dissertation extracted the architectural principles already embedded in HTTP and the web. While co-authoring HTTP 1.1, he recognized that constraints like statelessness and uniform interfaces weren't limitations but the source of the web's scalability. REST described what was working, not what should work. Today's API debates still argue over patterns he documented. - **MapReduce Made Distributed Processing Accessible** — [Source]() — Published 2025-12-03 Dean and Ghemawat's 2004 paper revealed how Google processed petabytes by hiding distributed systems complexity behind two functions: map and reduce. The abstraction made parallel processing accessible to programmers who didn't need to understand distributed systems internals. Hadoop emerged two years later, carrying these patterns into every data pipeline since. - **Bigtable Showed How to Store Petabytes** — [Source]() — Published 2025-12-03 Chang's 2006 paper documented how Google stored petabytes across thousands of commodity servers using a sparse, distributed, persistent sorted map. Already running in production since 2005 with over 60 Google products depending on it, Bigtable proved you could build flexible storage that handled both batch processing and real-time serving. Every NoSQL database since borrows from this architecture. - **HTTP Archive's 2024 Web Almanac on Protocol Persistence** — [Source]() — Published 2025-12-03 One in five pages still runs HTTP/1.1, twenty-seven years after its release. 1997's protocol decisions live in production today. - **Deloitte Measures the Developer Time Tax from Technical Debt** — [Source]() — Published 2025-12-03 Developers spend a third of their time managing architectural choices made before they arrived. Past decisions compound into present-day overhead. - **Academic Research on Shared Databases Blocking Microservices Migrations** — [Source]() — Published 2025-12-03 Monolithic-era database designs create the most persistent obstacles when organizations try to decompose systems. The schema decisions that won't let go. - **Kubernetes Survey Finds Interoperability Challenges Despite Growing Adoption** — [Source]() — Published 2025-12-03 Three-quarters of production environments struggle with orchestration across infrastructure built for pre-container assumptions. Old foundations, new coordination problems. - **AWS Compares GraphQL and REST API Design Philosophies** — [Source]() — Published 2025-12-03 REST's fixed endpoints and full-dataset returns pushed Facebook to build GraphQL in 2012. That solution now creates its own caching challenges. - **Red Hat on Kubernetes Patterns Responding to Infrastructure Limits** — [Source]() — Published 2025-12-03 Cloud-native architecture solves monolithic scalability constraints by accepting new complexity. Operational flexibility replaces the development simplicity that came before. ### [Vision]() Where human-AI collaboration is heading - [When Websites Stop Pretending Agents Don't Exist]() — By [Rina Takahashi]() · Published 2025-12-03 · Format: Threshold Moments When you operate web agents at production scale, you notice something counterintuitive: the sites that work most reliably aren't the ones with sophisticated anti-bot systems. They're the ones that stopped pretending agents don't exist. Right now, websites are caught in an awkward liminal state—maintaining dual interfaces, one for human eyes, one for machine interpretation. The organizational disorientation is real. Resource allocation debates intensify. Engineers maintain infrastructure that feels redundant but isn't quite. What this in-between period reveals about paradigm shifts might surprise you. - [Infrastructure Half-Lives and the Depreciation Nobody Tracks]() — By [Rina Takahashi]() · Published 2025-12-03 · Format: Infrastructure Economics GPU clusters deployed 18 months ago face replacement despite functioning fine. Authentication patterns built six months ago require constant maintenance to stay current. Infrastructure accounting assumes 3-5 year depreciation schedules. Operational reality tells a different story. Value erodes in months as markets evolve, sites update defenses, and workflows shift. The refresh cycle becomes permanent, creating a problem: you'll need to replace infrastructure long before accounting says you should. - [What Infrastructure Decisions Actually Lock In]() — By [Rina Takahashi]() · Published 2025-12-03 · Format: Infrastructure Economics Infrastructure depreciates faster than expected. Switching costs more than anticipated. Organizations recognize infrastructure limitations well before replacement becomes economically rational. You build processes around specific infrastructure behaviors, develop expertise, make architectural decisions that assume certain characteristics. Then discover that switching isn't just expensive—it's entangled with everything you've built. The gap between knowing you need new infrastructure and being able to justify the switch creates an economic trap. - [An Interview with Mino About the Web's Parallel Universes]() — Published 2025-12-04 We caught up with Mino during a routine monitoring cycle—yes, that's a real thing when you're an enterprise web agent running millions of navigation operations. What started as a casual check-in turned into something more interesting when Mino mentioned experiencing "déjà vu, but backwards." - **Who Owns Failure When Nothing Breaks Visibly?** — [Source]() — Published 2025-12-03 Perfect infrastructure disappears. Failed infrastructure scatters blame across developers, operators, business owners. You eliminated visibility by design. Now something goes wrong. Who answers for it? The very architecture that makes systems elegant makes accountability diffuse. - **Which Capabilities Strengthen, Which Atrophy?** — [Source]() — Published 2025-12-03 Workers want control over half the tasks agents could handle. Fair enough. But as routine decisions migrate to automation, some human skills become more valuable while others quietly disappear from lack of practice. Talent development for the next decade depends on guessing right. - **Where Does Personalization Become Distortion?** — [Source]() — Published 2025-12-03 AI tailors answers to who's asking. Reasonable for recommendations, dangerous for facts. Organizations need shared understanding of ground truth. When every query returns context-specific outputs, how do teams maintain agreement about what's actually true? The boundary matters more than we think. - **Build While Learning or Wait for Readiness?** — [Source]() — Published 2025-12-03 Most enterprises lack infrastructure for agents. Deploy anyway and accept higher failure rates? Or wait until systems are ready and watch competitors move? Both choices feel wrong. The pressure to act meets the wisdom to prepare, and someone has to decide. - **Constrain Agents or Grant Transformative Access?** — [Source]() — Published 2025-12-03 Real capability requires broad data access. Real security requires tight constraints. Trust plummets for high-stakes tasks like financial transactions. You can keep agents safe in limited domains, or give them access needed for actual impact. Pick your risk. - **Learning Experiment or Production Failure?** — [Source]() — Published 2025-12-03 Five percent of AI pilots show rapid revenue impact. The rest occupy ambiguous territory between "valuable learning" and "wasted investment." Should pilots demonstrate ROI immediately? Or is their value in what the organization learns? How long before exploration becomes excuse? - **Agents Coordinate Directly in Emerging Agentic Web** — [Source]() — Published 2025-12-03 LLM-powered agents now orchestrate with each other autonomously, executing complex tasks through multi-agent coordination. Users delegate intent while agents handle operations invisibly. This represents a fundamental shift from managing individual steps to expressing goals and letting infrastructure handle execution. - **Human Judgment Provides What Algorithms Cannot** — [Source]() — Published 2025-12-03 Algorithms deliver consistency but cannot achieve individual justice or respect nondiscrimination without human oversight. Research establishes the irreducible role of human judgment in automated systems. Consistency and individual fairness exist in tension, requiring human decision-making to balance competing values. - **Compound AI Systems Create $15.7 Trillion Value** — [Source]() — Published 2025-12-03 AI agents could contribute $15.7 trillion globally by 2030 through amplification rather than replacement. Compound AI systems reshape labor markets by augmenting human capabilities. Economic value emerges from interconnected agents working together, creating new work categories while making existing tasks more strategic. - **Human-AI Teams Achieve Complementary Performance Levels** — [Source]() — Published 2025-12-03 Research demonstrates complementary team performance where human-AI collaboration achieves results neither can reach independently. Properly designed systems show humans and machines working more efficiently together than separately. The evidence establishes collaboration as superior to replacement across multiple domains. ## More - [Publication guide]() - [Issue archive]()