--- title: "Issue 7" canonical: "https://current.tinyfish.ai/issue/7" publication: "TinyFish" publisher: "TinyFish" language: "en-US" guide: "https://current.tinyfish.ai/llms.txt" permalink: "https://current.tinyfish.ai/issue/7" issue: 7 description: "Infrastructure doesn't wait for permission—it forces decisions. This week: organizations restructuring before agents exist, frameworks collapsing into foundations, and the hidden costs of making reliability look easy." authors: ["Sable Whitford","Nora Kaplan","Rina Takahashi"] published: 2025-12-14 updated: 2026-11-05 articles: 92 --- # 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 7 Infrastructure doesn't wait for permission—it forces decisions. This week: organizations restructuring before agents exist, frameworks collapsing into foundations, and the hidden costs of making reliability look easy. ### [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. - [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. - [Your Agent's Questions Need a Recipient and a Timer]() — Published 2026-10-30 When an AI agent stops mid-task to ask a human something, it often reads as a failure of autonomy. In one study, letting coding agents ask about bug reports stripped of key details won back most of the lost ground, lifting solve rates from about 55% to 69.4%. But the "human" answering was another model, holding the full report and standing by. Real people are in meetings, or asleep. The design work is in the ask: who gets it, what it says, and what happens while it waits. - [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. - [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. ### [Market Pulse]() Reading the agent ecosystem through a practitioner's lens - [When Organizations Commit to Agents They Don't Yet Have]() — By [Nora Kaplan]() · Published 2025-12-10 · Format: The Signal The questions arrive after the commitment. How do we handle authentication across fifty regional variations? What happens when site structures change mid-workflow? How do we monitor behavior we can't see inside? Teams reach out saying they've adopted agents, then reveal through their requirements that they've restructured work around capabilities they're still building. The gap between organizational change and infrastructure reality is widening—and the teams caught in it can't go back. - [Framework Wars End Through Consolidation, Not Competition]() — Published 2025-12-10 The agent framework explosion of 2023-2024 is collapsing faster than anyone expected. Microsoft merged AutoGen with Semantic Kernel in October. LangChain pivoted from chains to LangGraph. CrewAI went from interesting to $18M Series A with 60% Fortune 500 penetration. - **Hybrid Workflows Beat Full Autonomy** — [Source]() — Published 2025-12-10 Stanford and CMU researchers measured autonomous agents at 88% faster and 90% cheaper than humans, but success rates dropped 32-50% on complex tasks. Hybrid systems routing structured work to agents while keeping judgment calls with humans improved performance 68.7%, revealing where autonomy actually delivers value. - **Modular Agents Enable Independent Optimization** — [Source]() — Published 2025-12-10 Plan-and-Act splits agents into separate Planner and Executor components, achieving 57.58% success on WebArena-Lite and 81.36% on WebVoyager. Using synthetic data for training, the framework lets you optimize planning and execution independently without rebuilding the entire system when requirements change. - **Direct Agent Communication Beats Centralized Routing** — [Source]() — Published 2025-12-10 Anemoi's semi-centralized architecture lets agents communicate directly rather than routing everything through a single planning LLM. Using GPT-4.1-mini as planner, it hit 52.73% accuracy on GAIA benchmark, beating the strongest open-source baseline by 9.09%. Better coordination protocols outperform throwing larger models at orchestration. - **Execution Constraints Belong In Planning** — [Source]() — Published 2025-12-10 REMAP integrates execution time estimation into multi-agent path finding, accounting for kinodynamic constraints, communication latency, and controller variability. Testing with 300 agents on benchmark maps showed 20% improvement in solution quality over baselines that treat execution as deterministic, making plans that survive contact with reality. - **Durable Execution Recovers From Stochastic Failures** — [Source]() — Published 2025-12-10 Traditional orchestrators expect deterministic steps. Agents break that assumption constantly. Purpose-built engines like Zipline, Inngest, and Hatchet combine queue, state store, and retry logic. They recover from failure points without expensive re-execution. Production viability depends on this infrastructure. - **Workflow-Level Economics Replace Billing Guesswork** — [Source]() — Published 2025-12-10 A single agent action fans out. Multiple model calls, tool invocations, RAG lookups. Each has different cost curves. Specialized tooling like Paid abstracts usage into workflow-level metrics. You get actual margin tracking instead of spreadsheet estimates. Financially essential, photographically boring. - **Holistic Benchmarks Test Beyond Accuracy Theater** — [Source]() — Published 2025-12-10 CLASSic evaluates cost, latency, stability, and security alongside accuracy. Enterprise deployment requires knowing if agents are prohibitively expensive, slow to respond, or easily exploited. Comprehensive frameworks test 500-1000 scenarios. They include failure modes, not just happy paths. - **Semantic Layers Prevent Hallucination Chaos** — [Source]() — Published 2025-12-10 LLMs accessing unorganized data lakes generate unreliable outputs. Structured catalogs and semantic layers define metrics, hierarchies, relationships. They give models the context needed for accurate queries. RAG shines when paired with this foundation, not raw data dumps. - **Memory Systems Make Agents Feel Adaptive** — [Source]() — Published 2025-12-10 Current-gen agents are stateless. Each task starts fresh. Emerging memory layers from Pinecone, Mem0, Zep handle context, conflict resolution, summarization with low latency. MCP approval hooks integrate human oversight with SLA timers. The difference between brittle and intelligent. - **Guardrails and Rollback Catch Failures Early** — [Source]() — Published 2025-12-10 Production reliability requires guardrails checking LLM outputs against criteria before downstream use. Retry mechanisms for failures. Prompt rollback enables version control and audit trails. Real-time monitoring reduces mean time to detection by 75%. Infrastructure that prevents user-facing disasters. - **Linux Foundation Bets on Open Standards for Agent Interoperability** — [Source]() — Published 2025-12-10 AAIF launch with MCP donation shows real architectural substance. Watch the infrastructure choices, not the announcement theater. - **IDC: Stop Treating Agentic AI Like Another Tool** — [Source]() — Published 2025-12-10 Traditional best practices create over-engineered platforms. The architectural archetype framing changes how you think about orchestration. - **Bloomberg Tests AI Browsers: Chrome Isn't Obsolete Yet** — [Source]() — Published 2025-12-10 Month of testing OpenAI and Perplexity browsers surfaces bugs and basic failures. Useful reality check on production readiness. - **How Orchestration Actually Works in Insurance Workflows** — [Source]() — Published 2025-12-10 Genre digs into observability and information domains for reinsurance. Finally, vertical-specific analysis that shows operational translation. - **Deloitte Data Exposes the Agent Readiness Gap** — [Source]() — Published 2025-12-10 Survey shows 80% claim automation maturity, only 28% for agents. ROI expectations reveal where enterprise confidence actually sits. - **McKinsey: 80% of Organizations Hit Agent Security Issues** — [Source]() — Published 2025-12-10 Research quantifies widespread risky behaviors. Skip the $4.4T projection, read the security-first deployment playbook instead. ### [Foundations]() Conceptual clarity earned from building at scale - [When to Automate and When to Keep Humans in the Loop]() — By [Nora Kaplan]() · Published 2025-12-10 · Format: Field Guide Teams come to us after watching agent demos, asking: "Can we automate our entire competitive monitoring process?" Sure. Technically, we can. That's exactly when things fall apart. The agents work fine. Trust collapses. Pricing data gets extracted incorrectly after a site redesign, and suddenly the whole system feels unreliable. Operating web agents across thousands of sites taught us this: technical capability isn't the constraint. Knowing where to draw the automation boundary is. - [When Scrapers Break Clean]() — By [Rina Takahashi]() · Published 2025-12-10 · Format: Tools in Context Your scraper stops returning prices overnight. Every field comes back null, and your monitoring catches it immediately because the data structure broke. This is the kind of failure teams can work with—loud, obvious, fixable. Rule-based validation exists for these moments, stopping bad data before it flows downstream. At scale, catching structural breaks isn't optional. It's survival. - [When Data Drifts Quietly]() — By [Rina Takahashi]() · Published 2025-12-10 · Format: Tools in Context Your scraper runs perfectly for months. Every field validates, types match, formats check out. Then someone in analytics notices the "product weight" field now contains shipping estimates. The data structure is fine. The meaning has drifted. This is what rule-based validation misses—the quiet corruption where everything looks right but nothing means what it should anymore. - [Procurement Teams Can't Buy What They Can't Measure]() — Published 2025-12-10 Watch what happens when procurement departments try to evaluate AI agents. They pull out the same RFP templates they use for CRM systems. Uptime guarantees. Fixed feature lists. Predictable outputs. - **Can You Retrofit Explainability Later?** — [Source]() — Published 2025-12-10 Teams skip explainability during POCs, planning to add it before launch. This almost never works. Explainability requires architectural decisions from day one. Skip it now and you're building technical debt that gets exponentially harder to fix. Ask how you'll explain decisions to users before you write the first line of code. - **Does Your Database Scale Sideways?** — [Source]() — Published 2025-12-10 When you need more capacity, can you add servers or just bigger servers? Horizontal scaling determines whether you grow or rebuild from scratch. Many databases make adding machines nearly impossible. Your initial choice locks you into a scaling path that's expensive to escape. - **What's Your SLA When Outputs Are Probabilistic?** — [Source]() — Published 2025-12-10 Traditional SLAs cover uptime and latency. AI systems need agreements around accuracy, adaptation speed, and decision quality. Cloud providers focus on availability because quality degradation resists contractual definition. If your SLA only measures uptime, you're missing what actually matters for AI. - **Does Your Observability Explain Why?** — [Source]() — Published 2025-12-10 Monitoring shows when something breaks. Observability shows what's happening and why it's happening. For agents, track context drift and reasoning patterns, not just errors and latency. Companies with structured evaluation frameworks see sixty percent fewer production incidents than those watching dashboards alone. - **How Different Is Production From Your Test Data?** — [Source]() — Published 2025-12-10 Offline evaluation uses curated datasets in controlled environments. Online evaluation monitors real user interactions. The gap between them reveals model drift, unexpected queries, and edge cases your test data missed. Testing only offline means optimizing for conditions that don't exist in production. - **Can You See Cost Per Request Live?** — [Source]() — Published 2025-12-10 An agent calling an LLM five times might be justified quality improvement or a bug causing runaway costs. Real-time cost visibility reveals whether you're trading money for quality or bleeding budget on inefficiency. Without live monitoring, expensive mistakes show up in monthly bills, not deployment dashboards. - [A Conversation with the Algorithm That Decides If You're Human Enough]() — Published 2025-12-10 We're in what Vee Three calls "the observation deck," though it's really just a conference room with too many monitors. Numbers flicker across every screen: 0.7, 0.3, 0.9, 0.1. Each one is Vee Three's verdict on whether someone, somewhere, is human enough. ### [Practitioner's Corner]() Lessons from the field—what we see building at scale - [What 98% Reliability Actually Costs]() — By [Nora Kaplan]() · Published 2025-12-10 · Format: Field Dispatch When a web agent platform reports 98% success rates across thousands of sites, that number looks clean on a dashboard. Behind it sits an on-call engineer staring at authentication failures, making a judgment call. Is this our proxy network or did these sites just rotate security certificates? - [When Your Auditor Doesn't Have a Category for Prompt Injection]() — By [Rina Takahashi]() · Published 2025-12-10 · Format: Builder Profiles An attacker needs only your email address to hijack your ChatGPT session, replace banking details in responses, extract data from Copilot. You never opened the malicious email. The attack happened anyway. Some vendors patched it. Others called it intended functionality. - [False Failures: Anywhere From 0% to 100%]() — Published 2025-12-10 When your automated tests fail, how many failures are real? Anywhere from none to all of them. - **Same Mistake, Three Weeks Apart** — [Source]() — Published 2025-12-10 Cloudflare's December 5 outage affected 28% of HTTP traffic. Root cause: routine config change. Same as November 18. Safeguards identified after first incident weren't deployed yet. "Would have helped prevent today's impact but, unfortunately, we have not finished deploying them." - **Trust Gap at Six Percent** — [Source]() — Published 2025-12-10 HBR surveyed 603 business leaders on AI agents. 6% fully trust them with core processes. 43% limit to operational tasks, 39% confine to supervised work. Security and privacy concerns dominate. Gap between spending and deployment keeps widening. - **Stability Overrides Published Schedules** — [Source]() — Published 2025-12-10 Atlassian began enforcing rate limits August 2025 but reserves right to enforce earlier "where apps are highly impacting platform stability." Limits factor app identity, tenant, request volume, product edition, user count. Burst and quota periods tracked independently. - **Anthropic Takes 40% Enterprise Spend** — [Source]() — Published 2025-12-10 Anthropic now captures 40% of enterprise LLM spend versus 24% last year, unseating OpenAI. Serves 300,000+ business customers, large accounts growing 7x. OpenAI declared "code red" December 2, pausing side projects to refocus on core experience. - **One in Five Reports Full Readiness** — [Source]() — Published 2025-12-10 20% of organizations say technology infrastructure fully supports agentic AI for core processes. 15% report data and systems ready, 12% have risk controls in place. Composite index: 27% leaders, 50% followers, 24% laggards. Spending outpaces preparation. - **APIs Now 80% of Web Traffic** — [Source]() — Published 2025-12-10 API traffic exceeds 80% of all web traffic per Akamai's State of the Internet report. 75% of API issues stem from mishandled limits. Modern APIs spanning multiple gateways create reliability problems when nodes track limits independently. - **How Raft Actually Works in Production Systems** — [Source]() — Published 2025-12-10 Consensus theory meets reality. Google, Netflix, Kafka implementations show what leader election costs at scale. - **Real Performance Numbers for Computer-Using Agents** — [Source]() — Published 2025-12-10 OpenAI's CUA hits 87% on navigation vs Claude's 56%. Benchmark data that matters for model selection. - **2024 Infrastructure Attack Patterns and Operational Impact** — [Source]() — Published 2025-12-10 DDoS surged 49%, major breaches at Finastra and Halliburton. Defense considerations when building automated systems. - **Where Engineering Orgs Actually Spend Resources in 2024** — [Source]() — Published 2025-12-10 ICONIQ data: 35-45% hiring AI roles, platform engineering priorities. Useful for benchmarking your own allocation decisions. ### [Echoes]() Past infrastructure decisions echoing in today's production realities - [The Backward Compatibility Tax]() — By [Rina Takahashi]() · Published 2025-12-10 · Format: Infrastructure Inflections Ken Thompson and Rob Pike designed UTF-8 on a placemat in a New Jersey diner in September 1992. The first 128 characters would map identically to ASCII. UTF-8 files using only ASCII would be byte-for-byte identical to ASCII files. Existing software could suddenly handle universal character encoding without modification. - [Docker's Lightness Created Monitoring Chaos]() — Published 2025-12-10 Docker launched in 2013 promising simplicity: package once, run anywhere. Containers were lighter than VMs, more portable, more efficient. Infrastructure teams adopted them fast. - **More Machines or Bigger Machines** — [Source]() — Published 2025-12-10 Every growth threshold forces the same split: horizontal for resilience, vertical for simplicity. Add servers, manage complexity. Upgrade hardware, accept single points of failure. The trade-off never changes. You're standing at this fork right now, same as everyone before you. - **Bandwidth Grows Faster Than Budgets Allow** — [Source]() — Published 2025-12-10 Global traffic reached 4.8 zettabytes in 2025. IT budgets grew 5% annually. The math doesn't work. Traffic scales exponentially while infrastructure investment crawls linearly. Your network congestion isn't an anomaly. It's the inevitable result of this widening gap playing out in production. - **Circuit Breakers Appear at Scale** — [Source]() — Published 2025-12-10 Small systems fail fast and restart. Large systems need circuit breakers, bulkheads, retry logic. The transition happens reliably: monolith works until traffic spikes, then you add the same resilience patterns everyone discovers. Microservices architectures rediscover them independently. You're implementing these patterns now. - **Technology Waves Compress Into Shorter Cycles** — [Source]() — Published 2025-12-10 Industry 1.0 took decades to spread. Industry 4.0 took years. AI infrastructure is scaling in months. Each wave compresses the timeline. Adoption curves steepen, investment cycles shorten. Small manufacturers implement Industry 3.0 tools while hyperscalers deploy next-generation AI. The gap widens continuously. - **Single Cloud Becomes Hybrid Infrastructure** — [Source]() — Published 2025-12-10 Start on one cloud for speed. Scale reveals vendor lock-in risk. Add a second cloud for resilience. Now you're managing hybrid complexity. 46% of enterprises made this exact transition in 2024. Centralize for simplicity, distribute for survival, manage the operational overhead forever. - **Queues Replace Real-Time Processing Under Load** — [Source]() — Published 2025-12-10 Systems start synchronous. Every request gets immediate response. Load increases, latency spikes. Queues emerge: decouple request from processing, smooth the peaks. Queue-based load leveling appears reliably when real-time breaks. Your architecture is discovering this solution right now, same as everyone before you. - **Paxos Made Distributed Consensus Possible** — [Source]() — Published 2025-12-10 Leslie Lamport's consensus algorithm sat in a filing cabinet from 1989 to 1998 before publication. It proved distributed systems could agree on state despite failures. Every database claiming "distributed" today inherits Paxos's fundamental trade-offs between consistency and availability. - **Chord Built the Foundation for Peer-to-Peer Systems** — [Source]() — Published 2025-12-10 MIT's 2001 distributed hash table protocol showed how to locate data across thousands of nodes with logarithmic lookups. Chord's consistent hashing and finger tables became the blueprint for BitTorrent, content delivery networks, and systems that distribute data without central coordination. - **Google File System Normalized Failure as Infrastructure Reality** — [Source]() — Published 2025-12-10 Google's 2003 paper described a file system assuming hardware fails constantly. GFS's design proved commodity hardware could store petabytes if you accepted failure as normal, not exceptional. HDFS and S3 followed this blueprint of weak consistency and single-master architecture. - **Dynamo Made Eventually Consistent Systems Practical** — [Source]() — Published 2025-12-10 Amazon's 2007 paper revealed how they built a key-value store that stayed available during failures by accepting temporary inconsistency. Dynamo's techniques spawned Cassandra, Riak, and the NoSQL movement. Shopping carts mattered more than perfect consistency. - **Colorado's $465 Million Technical Debt Crisis** — [Source]() — Published 2025-12-10 Decades of "just one more patch" thinking on aging systems now demands half a billion in deferred infrastructure costs. - **Atlassian's Microservices Velocity Problem** — [Source]() — Published 2025-12-10 Distributed architecture built atop monolithic data models: today's deployment friction reflects yesterday's database coupling assumptions. - **Martin Fowler on Database Schema Evolution** — [Source]() — Published 2025-12-10 Why schema changes still break everything: waterfall-era "design once, never touch" philosophy baked immutability into relational databases. - **Multi-Cloud Identity's Security Debt** — [Source]() — Published 2025-12-10 IAM modernization pain traces to early cloud decisions that bolted authentication onto existing systems rather than rebuilding foundations. - **IDC: 82% Need Cloud Rework After Migration** — [Source]() — Published 2025-12-10 Lift-and-shift strategies preserved on-premises architectural patterns wholesale, creating cloud systems that inherited rather than resolved legacy constraints. - **Salesforce and Twilio's Decade-Old API Versions** — [Source]() — Published 2025-12-10 Early REST design choices created permanent backward compatibility contracts: one interface decision becomes infrastructure you maintain forever. ### [Vision]() Where human-AI collaboration is heading - [When Authentication Moved to Hardware]() — By [Rina Takahashi]() · Published 2025-12-10 · Format: Threshold Moments For years, when a website tightened its defenses, we'd adapt. Better session management. Smarter timing. More sophisticated patterns. Code could always get better. - [Infrastructure's Unavoidable Tax]() — By [Rina Takahashi]() · Published 2025-12-11 · Format: Infrastructure Economics Infrastructure teams learn observability's cost structure the hard way: it consumes 10-30% of budgets, often landing at the high end. What begins as $50,000 for a monolithic application becomes $14 million for distributed systems—not from business growth, but architectural evolution. Each microservice multiplies telemetry volumes. Each autoscaling event creates permanent billing increases. By the time the trajectory becomes visible, the infrastructure already depends on it. The economics aren't negotiable anymore. - [When Observability Pays for Itself]() — By [Rina Takahashi]() · Published 2025-12-11 · Format: Infrastructure Economics That 10-30% observability spend creates a sharp question: what does it actually buy? Some organizations achieve 302% ROI through faster incident response and freed engineering capacity. Others accumulate costs without corresponding value. The gap isn't tools—it's whether infrastructure depth enables decisions that wouldn't happen otherwise. When observability turns telemetry into outcomes, the economics work. When it just generates data nobody uses, the spend becomes pure overhead. - [Interview with Mino: The Web Is Becoming Eerily Similar Under the Hood]() — Published 2025-12-10 We're sitting in what Mino calls "the observation deck"—though it's really just a conference room with a wall of monitors showing real-time web traffic patterns. Mino is an enterprise web agent that navigates thousands of sites daily, and they've been noticing something strange lately. - **Who Owns Decisions Nobody Made?** — [Source]() — Published 2025-12-10 An agent acts autonomously and causes harm. The developer built general capabilities. The company deployed it. The user set parameters. Everyone contributed, nobody decided. Accountability assumes someone chose—but autonomous systems dissolve that assumption. Where does responsibility land when it can't map cleanly to any single actor? - **What Breaks Before Anything Actually Breaks?** — [Source]() — Published 2025-12-10 AI infrastructure doesn't fail with alarms and error messages. Quality erodes. Responses drift. Confidence scores slip. Nothing technically breaks, so nothing triggers alerts. By the time you notice, the damage has compounded across teams and decisions. What monitoring catches problems that announce themselves through gradual wrongness? - **Does Truth Exist Outside Individual Experience?** — [Source]() — Published 2025-12-10 Every user gets information shaped by their preferences, history, and inferred beliefs. Your reality diverges from mine by design. When we need to make collective decisions—as organizations, as societies—what happens if we've lost the shared foundation that makes agreement possible? - **What Muscles Atrophy When Routine Disappears?** — [Source]() — Published 2025-12-10 Agents handle preparation, research, drafting. Humans curate and decide. Sounds efficient. But expertise builds through repetition of fundamentals—the boring work we're automating away. When we skip the reps that build foundational capability, what exactly are we developing the judgment to evaluate? - **Can Security Frameworks Handle Self-Directed Software?** — [Source]() — Published 2025-12-10 Enterprise security assumes every action traces to a person with credentials and permissions. Agents don't fit. They learn, adapt, make decisions across contexts. Do we retrofit them into identity systems built for humans, or admit those systems can't contain what agents actually do? - **Lock Down Standards or Learn Through Deployment?** — [Source]() — Published 2025-12-10 Strong governance before deployment prevents chaos but may prevent learning what agents actually need. Deploy first and you discover real requirements, but risk uncontrolled sprawl. Capabilities evolve faster than frameworks. When's the right moment to establish guardrails that won't immediately become obsolete? - **Skill Partnerships Reshape Work in AI Era** — [Source]() — Published 2025-12-10 McKinsey research shows AI could automate 57% of US work hours, yet employment keeps growing as new skill partnerships emerge. Companies now hire agent product managers and AI validators. Mayo Clinic expanded radiology staff 50% since 2016 while deploying hundreds of AI models. - **Worker Desires Misalign With AI Investment** — [Source]() — Published 2025-12-10 Stanford tracked 1,500 workers across 104 occupations. Result: 70 million US workers face major transition, with 41% of Y Combinator AI investments concentrated in low-priority areas. Critical mismatch between worker needs and current agent development suggests human competencies are shifting toward interpersonal skills. - **When to Automate Versus Augment With AI** — [Source]() — Published 2025-12-10 Management Science research establishes framework showing optimal AI use depends on complementarity types. Between-task complementarity favors automation, within-task complementarity favors augmentation. Framework reveals how AI benefits shift based on whether tasks are performed independently or collaboratively with humans. - **Agentic Web Framework Maps Agent Infrastructure Future** — [Source]() — Published 2025-12-10 July 2025 research establishes three-dimensional framework for understanding how AI agents reshape web infrastructure: intelligence, interaction, economics. Framework addresses who acts on the web, how tasks execute through delegated orchestration, and what the web produces beyond content. ## More - [Publication guide]() - [Issue archive]()