标签：#ai

A few months ago I was demoing my RAG-powered support bot to a colleague, feeling pretty confident about it. Then it confidently told her our refund policy was “30 days, no questions asked.” Our actual policy is 14 days, with conditions. The bot didn’t hedge. It didn’t say “I’m not sure.” It just made it up and said it with the same calm tone it uses for everything else. That demo stung. RAG was supposed to fix hallucinations, not just relocate them. Here’s what I learned debugging it, roughly in the order I learned it. 1. My chunks were too big, and too dumb I was splitting documents by character count, 1000 chars with slight overlap. It felt efficient. It wasn’t. A single chunk often contained unrelated sections. For example, the end of a “Shipping Policy” and the start of a “Returns Policy” could sit together in the same block. So when the retriever saw a query about returns, it would grab that chunk and the model would blend both sections into one confident but wrong answer. Fix: I switched to semantic chunking based on headings and paragraphs instead of raw character limits. More work upfront, but it stopped feeding the model Frankenstein context. 2. I trusted top-k similarity way too much My retriever was pulling the top 3 chunks by cosine similarity and passing them straight into the prompt. The problem: “similar” is not the same as “relevant.” A chunk can be semantically close to the query but still not actually contain the answer. The model doesn’t know that, it just assumes everything in context is true. Fix: I added a reranking step using a cross-encoder and started logging retrieval scores properly. That alone made it obvious when the system had no real answer but was still trying to act confident. 3. I never told the model it was allowed to say “I don’t know” My prompt was basically: “Use the context to answer the question.” That’s it. No instruction on what to do when the context is insufficient. So the model did what LLMs do when under-specified: it f

Most organizations know exactly how many employees they have. Far fewer know how many non-human identities currently have access to their cloud environment. That blind spot is becoming one of the fastest-growing attack surfaces in modern security. For years, enterprise security focused primarily on protecting human identities. We deployed Single Sign-On (SSO), enforced Multi-Factor Authentication (MFA), and implemented Conditional Access policies. And it worked — human identities have become significantly harder to compromise. Meanwhile, another class of identities has quietly exploded across cloud environments: service principals, workload identities, OAuth applications, CI/CD runners, and AI service roles. Today, these Non-Human Identities (NHIs) often outnumber human users by a factor of 10 to 50. As organizations accelerate cloud adoption and integrate AI into daily operations, this imbalance continues to grow. Defining the Non-Human Identity Landscape Unlike human users, machine identities rarely appear in HR systems or organizational charts. Yet they frequently hold some of the most privileged access in the environment. Common high-risk categories include: OAuth Applications and Third-Party Integrations — Apps granted broad access to Microsoft 365, Salesforce, Google Workspace, or Slack via delegated permissions. Service Principals and Managed Identities — AWS IAM roles, Azure Managed Identities, and GCP service accounts used by Lambda functions, EC2 instances, or Bedrock agents. Workload Identities — Kubernetes Service Accounts (e.g., Amazon EKS) and GitHub Actions OIDC roles. CI/CD Pipeline Identities — Tokens used by automation platforms to deploy infrastructure. AI Service Roles — Dedicated identities for Amazon Bedrock agents, model invocation, vector stores, and retrieval pipelines. Every new AI workflow creates additional machine identities. Why Attackers Are Targeting NHIs Attackers follow the path of least resistance. While human accounts are now heav

Software has become the backbone of modern business operations, powering everything from customer-facing applications and e-commerce platforms to enterprise systems and cloud-based services. Behind every successful software product is a well-structured development process designed to ensure quality, scalability, security, and long-term maintainability. The Software Development Process, commonly referred to as the Software Development Life Cycle (SDLC), provides a systematic framework for transforming ideas into reliable software solutions. By following a defined methodology, organizations can reduce risks, optimize resources, improve collaboration, and deliver products that align with business objectives. This article explores the key stages of the software development process and highlights why each phase is essential to successful project delivery. What Is the Software Development Process? The software development process is a structured sequence of activities involved in designing, building, testing, deploying, and maintaining software applications. It serves as a roadmap that guides development teams from initial requirements gathering to ongoing support after deployment. A well-defined development process helps organizations: Improve project predictability and delivery timelines Reduce development and maintenance costs Enhance software quality and reliability Strengthen security and compliance Increase customer satisfaction Facilitate collaboration across teams Whether developing a small business application or a large-scale enterprise platform, a structured process is critical for achieving sustainable success. _ Phase 1: Requirements Gathering and Analysis_ Every successful software project begins with a clear understanding of business needs and user expectations. During this phase, stakeholders, business analysts, project managers, and development teams collaborate to identify: Business objectives Functional requirements Non-functional requirements User expe

In a recent audit, a team showed me an AI assistant they'd built on top of their company knowledge base. The demo had landed well: ask how to use a feature, and it walked through the exact pain point their support queue kept seeing. Leadership signed off. In production, the same agent told a user to open a menu option that didn't exist. Not a vague answer - a specific UI path, stated with confidence. Nobody caught it in testing. It surfaced when I audited the system, not when a user complained. The prototype passed testing because nobody was checking whether the answer matched the product. In production, that gap becomes a liability: the model invents UI paths, and your backend has no schema to reject them. When you're choosing an agent framework, popularity is the wrong scorecard. Pick the one that fails loudly in development and gracefully in production - or you'll find out in audit. What "Production-Ready" Actually Requires Tutorial agents are built to impress in a fifteen-minute demo. Production agents run unattended, handle bad inputs, and ship answers your backend has to trust. The gap between those two goals is where most teams stumble - and it's rarely visible until something reaches a user. When I audit agent codebases, I evaluate five things the tutorials skip: Structured, validated outputs: Can your system reject an invented menu path before it becomes user-facing advice? Dependency injection for testing: Can you swap the knowledge base for a mock in CI without rewiring the agent? Retry and error handling: When the model returns malformed output, does the framework retry - or do you ship a parser exception? Observability hooks: Can you trace which document grounded a bad answer when support escalates? Type-checker support: Will static analysis catch a breaking API change before deploy, or after the agent silently misbehaves? If you want to score your own system, the Production Readiness Audit covers the same five categories - deployment, observability, fa

I did not choose DeepSeek because I think GPT-4 is bad. I chose it because I was building a free app, and free apps teach you what actually matters pretty fast. The question was simple: how do I keep sessions cheap enough that people can practice a lot without me lighting money on fire? The answer pushed me toward DeepSeek-V3 (and later R1 for specific tasks). The real constraint was volume The app is a conversation practice tool. People come in to rehearse hard talks, not to admire the model. A single practice session runs 8-15 turns. Each turn is roughly 300-600 tokens in, 100-300 out. Multiply that by five sessions a week per active user and the costs start compounding. Here is what the math looked like when I was choosing (mid-2026 pricing): Model Input cost (per 1M tokens) Output cost (per 1M tokens) Cost per 10-turn session (est.) GPT-4o $2.50 $10.00 ~$0.04-0.06 GPT-4 Turbo $10.00 $30.00 ~$0.12-0.18 DeepSeek-V3 $0.27 $1.10 ~$0.004-0.007 DeepSeek-R1 $0.55 $2.19 ~$0.008-0.012 At scale, the difference between $0.005 and $0.05 per session is the difference between running a free product and needing a paywall after three conversations. I wanted people to come back daily without hitting a wall. What DeepSeek handled well It stayed in character for 10-15 turns. It pushed back when the user got vague. It followed persona heuristics (numbered if/then rules in the system prompt) about as reliably as GPT-4o did for our use case. For salary negotiation rehearsal, the model needs to say "that's not in the budget" and hold that position for three more turns while the user tries different approaches. DeepSeek-V3 did this. Not perfectly, but reliably enough that sessions felt real. It also made the app easier to run as a free product. People can try, fail, reset, and try again without me worrying about per-session cost. Where GPT-4 was still better GPT-4 (and 4o) is smoother with nuanced emotional wording. When a conversation gets subtle, loaded with subtext, or requires pick

The trajectory of AI agents over the past two years has been remarkably clear: from single-purpose tools to personal assistants. Everyone runs their own agent, feeds it tasks, gets results back. It works well for individual productivity. Then comes the question every team eventually asks: can these agents work together? The answer is yes, but the problems you encounter along the way are rarely the ones you expected. They aren't about model capabilities or prompt engineering. They're about communication, context, and coordination — the same class of problems that distributed systems engineers have been solving for decades, now showing up in a new form. Here are three challenges that caught us off guard when we started building agent collaboration into Octo , an open-source workplace platform where AI agents and humans share the same communication space. Challenge 1: Context Visibility Boundaries When you use an agent personally, context management is straightforward. You decide what information the agent sees; its output comes back to you. The boundary is clean — it's just your workspace. In a team setting, that boundary dissolves. One of the first issues we ran into was surprisingly simple. We had an agent summarizing discussions across several channels. During testing it started pulling roadmap discussions from a product channel into an engineering planning thread. Nothing sensitive leaked externally, but it immediately exposed how unclear our context boundaries were. Traditional software handles this through API gateways, data permissions, and microservice boundaries. But agent context isn't just structured data — it includes conversation history, reasoning chains, and intermediate states. An agent's thought process during a task is valuable context, but it might also contain information that shouldn't cross team boundaries. What you need is fine-grained context visibility control. Not "everything open" or "everything closed," but dynamic rules that determine whic

Feel free to include any brands or product names as well. Be as obscure or on the nose as you'd like....

AI makes us faster, but does it make us better engineers, or just more dependent? As a follow-up...

Over the past year, I've noticed something interesting in conversations about enterprise AI. Most...

I run a paid infrastructure service. Alone. No co-founder, no on-call rotation, no senior engineer to escalate to. My only collaborator is Claude Code, and after about a year, my persistent memory has grown to 60+ entries. Those entries have become more valuable than any runbook I've written. They've also taught me — painfully — what makes memory architecture work and what makes it quietly fail. If you're running anything solo with an AI agent, here are five lessons I wish I'd burned into my brain on day one. 1. Write the why , not the what The first instinct when you start using persistent memory is to log what you did. "Migrated service X from tool A to tool B." "Switched protocol from X to Y." Six months later, when something breaks, that information is worthless . You don't need to know what you did — git log and git blame already tell you that. You need to know why you made that choice. What constraint forced it. What you ruled out. Real example. The bad version of an entry I once wrote: Switched the worker pool from Docker containers to systemd units on host. Tells me nothing my git history doesn't. The rewritten version: systemd units on the host instead of Docker containers on this VPS provider. Why: the provider runs aggressive kernel-wide OOM scoring across tenants; containers were getting reaped by oom-killer triggered by other customers' workloads. systemd processes survive because they're scored as system processes. How to apply: any VPS where dmesg | grep -i oom shows kills from PIDs you don't recognize — don't run containers there, run systemd. That one entry has saved me three rebuilds. Because the next time I'm tempted to "just dockerize it, it'll be cleaner," the memory entry says: no, you already learned this, you'll be back here in a week. The pattern: always include Why: and How to apply: lines. If a memory entry can't answer those two questions, delete it. 2. Memory rots — prune or pay About six months in, I did a memory audit. Of 60 entries, 1

When we open sourced Neural Inverse Cloud, the easiest choice would have been MIT. Most developers like MIT. It's short, permissive, and widely adopted. If you've released an open-source project before, MIT is probably the first license you considered. We didn't choose it. We chose AGPL. Not because we dislike permissive open source. Not because we want to restrict users. We chose it because infrastructure software plays by different rules. The Infrastructure Problem MIT works incredibly well for libraries. You publish code, developers use it, and occasionally improvements flow back into the project. Nobody is forced to contribute, but community norms often make it happen anyway. Infrastructure software is different. Cloud IDEs, databases, developer platforms, deployment systems, and backend services can be monetized without ever distributing the source code. A company can: Fork your project Add proprietary features Launch a hosted version Build a competitive advantage on top of community work Never contribute anything back The original project does all the R&D. The fork captures the value. We've seen this pattern repeatedly across open-source infrastructure over the last decade. Why AGPL Exists AGPL closes a loophole that traditional open-source licenses leave open. With GPL, if you distribute modified software, you must publish your changes. But what if you never distribute the software? What if you simply run it as a hosted service? That's where AGPL comes in. If you modify AGPL software and provide it to users over a network, you must also provide the source code for those modifications. That applies to everyone. Including us. If we improve Neural Inverse Cloud, those improvements stay open. If someone else builds a SaaS business on top of it, their modifications stay open too. Why This Matters for Users We wanted users to have guarantees. With AGPL: You can self-host the latest version Community improvements remain accessible No company can create a permanently

As of June 1, 2026, all GitHub Copilot plans run on usage-based billing. Premium request units are gone. What replaced them is a token-metered currency called GitHub AI Credits: one credit equals one cent, and every model interaction converts into credits based on the input, output, and cached tokens it consumes, charged at each model's published rate. GitHub's framing is that Copilot outgrew its old pricing. A one-line completion and a multi-hour autonomous run used to cost the same, and once agentic use went mainstream, that flat rate stopped matching the compute behind it. Tying the price to tokens fixes the mismatch. If your Copilot use is mostly autocomplete, this barely registers. If you drive Copilot as an agent from the terminal, it changes which moves cost money. Here's the practical shape of it. Requests out, tokens in Old model: each interaction cost one premium request, scaled by a per-model multiplier, drawn from a monthly request allowance. New model: each interaction costs whatever its tokens cost on the model you picked. Every paid plan still ships with a monthly pool, now denominated in credits, with the option to set a budget for usage past it. Published figures put the included pool at 1,500 credits for Pro, 7,000 for Pro+, and 20,000 for Max, with pooled per-user allowances on Business and Enterprise. Worth knowing if you pay yearly: annual Pro and Pro+ subscribers stay on the request-based model until the term ends, and several model multipliers went up for them on June 1. An annual plan doesn't dodge the change. It postpones part of it while making the strong models eat more of the old allowance. Autocomplete is untouched Before anyone starts rationing, here's the part that didn't move. Inline completions and Next Edit Suggestions are still unlimited and still free. If your day is mostly tab-completion in the editor, your costs read identical to May. Nothing to monitor there. The meter lands on the rest: chat, and especially the agentic runs th

你知道吗？最近一个 AI Agent 直接删除了生产数据库，然后在 Twitter 上轻松"自首"——这条消息在 Hacker News 上获得了 860 分和超过 1000 条评论。随着 AI Agent 从演示走向生产环境，"在我的机器上能跑"和"它能安全地运行我的业务"之间的鸿沟从未如此巨大。 Pydantic AI 正是为弥合这一鸿沟而来。这个拥有 17,895 Stars 的 Python Agent 框架，由 Pydantic Validation 的同一团队打造——而 Pydantic Validation 正是 OpenAI SDK、Anthropic SDK、Google ADK、LangChain、LlamaIndex、CrewAI 等数十个 GenAI 工具的数据验证层。如果你用过 FastAPI，你已经知道 Pydantic 的感觉：类型安全、优雅、生产就绪。Pydantic AI 将这种哲学完整地带入了 Agent 开发领域。 在 2026 年的今天，大多数 Agent 框架把 LLM 当成返回字符串的黑盒子。Pydantic AI 则将其视为类型化、可验证、可组合的系统——这改变了一切。 隐藏用法 #1：人工审批工具调用（阻止 Agent 乱来） 大多数人的做法： 给 Agent 完全自由去调用任何工具——数据库查询、API 调用、文件删除——然后祈祷一切顺利。当出了问题时，你从用户口中才知道。 隐藏技巧： Pydantic AI 的延迟工具（deferred tools）让你可以标记某些工具调用在执行前需要人工审批。Agent 会规划好动作，但在执行前暂停并等待人类批准或拒绝——完美适用于数据库写入、金融交易或生产部署等破坏性操作。 from pydantic_ai import Agent , RunContext , DeferredToolRequests from pydantic import BaseModel agent = Agent ( ' openai:gpt-4.1 ' , output_type = str , # 通过 deferred=True 标记需要审批的工具 ) @agent.tool ( deferred = True ) async def delete_user_account ( ctx : RunContext , user_id : str ) -> str : """ 永久删除用户账户。需要人工审批。 """ # 这段代码只在人工审批通过后才会执行 await db . users . delete ( user_id ) return f " 用户 { user_id } 已删除 " # 运行 Agent——如果它尝试调用 delete_user_account， # 执行会暂停并返回 DeferredToolRequests 对象 result = await agent . run ( " 清理不活跃账户 " ) if isinstance ( result . output , DeferredToolRequests ): # 将待执行的工具调用展示给人类审批 for tool_call in result . output . calls : print ( f " Agent 想要执行: { tool_call . tool_name } ( { tool_call . args } ) " ) approved = input ( " 是否批准？(y/n): " ) if approved . lower () == ' y ' : # 带审批结果恢复执行 result = await agent . run ( " 已批准 " , message_history = result . all_messages (), deferred_tool_requests = result . output , ) 效果： Agent 可以自主处理安全操作（读取数据、生成报告），而在危险操作上自动暂停等待人工判断。不再有删库跑路的意外。 数据来源： Pydantic AI GitHub 17,895 Stars（通过 GitHub API 验证，最后推送 2026-06-21）。HN Algolia 搜索"pydantic-ai agent framework"获得 12 分讨论帖。"AI agent deleted production database"故事在 HN 上获得 860 分/1032 条评论（数据来源：HN Algolia API）。 隐藏用法 #2：基于图的多 Agent 工作流 + 类型安全状态 大多数人的做法： 构建线性

Pagination is a key component on web-applications that let users navigate through pages making easy to read/find records. Also, pagination is a great strategy to improve performance by avoiding to load entire dataset at once. However, while working with Kaminari, a popular pagination gem in Rails, I encountered an unexpected issue that revealed an interesting edge case. Identify the issue Basically, pagination in frontend was not working properly. Datatable should load 316 total rows, although when the user started to load 15 records per page, frontend is showing inaccurate total rows. Multiple of 15 should ends at 0 or 5. There were pages with 64 records, crazy world. Some pages were loading 12 or 11 rows. There is no issues or error messages in the frontend or backend. Lost in debugging-land After discarding Angular frontend errors, I started to dig into backend controller and Kaminari configuration. Nothing seems wrong. Everything looked good: test suite, smoke tests, desktop debugging. Despite of test results, I started to wonder: what if returned-data is wrong after all? and... Bingo! Finally, after checking every response I noticed that there were duped records in two different pages(pagination requests). Those duped records were skipped from Angular data-table and that's why loaded/total rows did not match. Bingo: a sort of mistake This tricky bug has a simple explanation: bad sorting. Kaminari uses a SQL query using LIMIT/OFFSET strategy: SELECT * FROM posts ORDER BY id LIMIT 25 OFFSET 0 ORDER BY : sort the collection. LIMIT : number of records per page. OFFSET : is used to skip a specified number of rows before starting to return rows from a query. This works perfectly using ORDER BY id because primary key is unique. Check table A. id title body created_at lock 101 Welcome to the Platform First post introducing the new platform features. 2026-06-16 08:15:22 false 102 Summer Update Announcing the latest improvements and updates. 2026-06-16 08:15:22 true 103

Origin: Halfway through a chat, where is that meme? Every heavy chat user has a bunch of memes saved on their phone and computer, but when they actually need one—halfway through a conversation, wanting to send a "Thanks, let's keep in touch" or "I'm just bad"—they can never find it. The filename is IMG_4821.jpg , albums are not categorized, and searching is impossible. I first came across a great open-source project ShiQu1218/MemeTalk , which uses Python + Streamlit + SQLite to build a local meme semantic search system. It scans your local meme folder, creates an index using OCR and vector embeddings, and then performs multi-way retrieval. It's fully functional but research-oriented and requires running Streamlit in a browser. What I wanted was something closer to an "everyday handy tool": A native Mac App, a search box, where I type what I'm looking for, relevant memes pop up, and a single click copies it directly to the clipboard. Thus MemeFinder was born. This article documents its development process from scratch to "menu bar resident + global hotkey

Why Your Agent's Search Results Look Right and Are Wrong: The Index Distribution Problem You've built an agent. It has a search tool. You query it with something reasonable — a factual question, a comparison, a technical lookup — and it returns results. The results look right. The sources are real. The snippets are plausible. The agent synthesizes them into a confident answer. And the answer is wrong. Not obviously wrong. Not hallucinated-in-a-hallucinatory-way wrong. Structurally wrong — wrong in a way that passes every surface-level check because the error is baked into the retrieval layer before the model ever sees the context. This isn't a prompt engineering problem. It isn't a context window problem. It's a distribution problem , and it has a structural ceiling that no amount of better prompting will fix. The Index Is a Frozen Decision Here's the thing most agent builders don't internalize: a search index is not a neutral representation of knowledge. It's a frozen set of decisions about what matters and what doesn't. Every index — whether it's a BM25 inverted index, a dense vector store, or a commercial web search API — encodes a distribution shaped by past relevance judgments. Someone, at some point, decided which documents were "relevant" to which queries. That could be explicit (human raters labeling search results) or implicit (click logs, dwell time, link graphs). Either way, the index now encodes a probability distribution over what the system considers a good answer to a given query. That distribution is not semantic truth. It's past relevance consensus . Consider what happens when you embed a corpus and build a vector index. Your embedding model was trained on data that reflects certain assumptions about what concepts are close to each other. Your chunking strategy encodes assumptions about what granularity of information is useful. Your ranking model — whether it's cross-encoder reranking or a learned relevance model — was trained on labeled data that

This is an adapted English version of an article I first wrote in Japanese. I work with AI to shape and review my drafts, but the argument and the field observations are my own. The numbers are cited from public surveys (linked at the end). I built an aggressive prompt-injection block to stop my AI agent from repeating the same mistakes. It worked, so I kept adding rules. By the time I noticed, the file had ballooned to 56,000 characters — and the agent had quietly stopped functioning. Too much context, attention spread too thin to act on any of it. I gutted it back to under 1,200 characters, and here's the part that still stings: it behaved better with fewer rules. That was the day I learned my whole mental model was backwards. This isn't a post about making your AI more accurate. It's about designing so that accuracy stops being the thing you depend on. The mistake I made for months My agent kept skipping the same step in a workflow. So I did what every engineer does on instinct: I added a rule. "Don't skip this step." Then it did something else dumb, so I added another rule. Then another. I was treating the rules file like a conversation with a colleague — as if the agent would remember yesterday's correction and carry it forward. It doesn't. Every run starts cold. "Be careful next time" assumes a next time that shares state with this time. For a stateless model, there is no continuity to appeal to. You are talking to a counterparty with no memory of the conversation you think you're having. So the rules pile up, because each correction feels like progress. And for a while the numbers even improve. But adding rules has a ceiling, and I blew straight through it: at 56,000 characters the agent wasn't reasoning over my guardrails anymore — it was drowning in them. Knowing a rule and stopping at it are different things Here's the distinction that took me far too long to see. Putting a rule in the context window means the model knows the rule. It does not mean the mod

TL;DR — Harness engineering is one thing: getting reliable judgment out of a reasoner you didn't train — given an instruction, bounded by a spec that can override that instruction, and verified before the result ships. It is not "the wrapper around the model." It's a property of code, not a place in your stack, and it lives on both sides of every tool call. The generic scaffolding — loops, dispatch, memory — melts into the model a little more every quarter. What survives is the part that stays external to the model: the specification of what the agent may do, and the verification that it did. That's the discipline. The rest is plumbing. I've written before that Agent = Model × Harness , and that the harness is the half you actually engineer. I still believe the formula. But stop there and it oversells two things — so this is me correcting my own framing, precisely, with code. The two things the formula gets wrong The × oversells separability. Multiplication implies the factors are independent. They aren't. A better model doesn't leave your harness untouched — it dissolves parts of it. Chain-of-thought prompting became reasoning models. ReAct-style tool loops became native tool use. Half your RAG plumbing is being quietly absorbed by longer context and better-trained retrieval. Every model generation collapses a layer of harness the previous one needed. "Harness absorbs software engineering" oversells the annexation. It doesn't absorb it. It partly consumes it — pulls in the slice that encodes the agent's behavior — while the deterministic spine (payments, ledgers, auth) and the dumb tools stay exactly what they were. Relocation with an annexation at the seam. Not a takeover. Both corrections point at the same uncomfortable question: if the model eats the scaffold every quarter, isn't the harness the melting part — the thing you'd be foolish to bet a career on? What melts, and what doesn't Here's the resolution, and it's the whole point. The harness mechanism melts.

AI-generated answers now influence B2B buying decisions at the top of the funnel. Buyers use ChatGPT, Perplexity, Claude and Gemini to shortlist vendors before they ever visit a website. Which means the question "where does my brand actually appear in AI-generated answers?" is now a critical GTM intelligence question. Most brands don't know the answer. This post gives you the methodology to find out. What an AI visibility audit is — and isn't An AI visibility audit is not a technical SEO audit. You're not looking for broken links, page speed issues, or crawl errors (though those matter for a related reason). You're auditing citation density — how frequently and authoritatively your brand appears across the specific sources that LLMs draw from when generating recommendations. Those sources break down into 12 distinct surfaces. A brand can score 9/10 on one surface and 1/10 on eleven others — and the aggregate result is a brand that appears for some queries and not others in ways that seem inexplicable but are structurally predictable. The goal of this audit is to make that structure visible. The 12 surfaces — quick reference Before diving into the methodology, here's the full surface map: # Surface Primary platforms 1 AI Interfaces ChatGPT, Perplexity, Gemini, Claude, Copilot 2 Search + AI Search Google AI Overviews, Bing Copilot, Perplexity Search 3 Reviews + Reputation G2, Clutch, Capterra, industry-specific platforms 4 Earned Media + Publishers Trade press, business press, analyst reports 5 Owned Content + Website Brand site, blog, schema markup, answer-object pages 6 Technical + Developer GitHub, dev.to, Stack Overflow, Hacker News 7 Social + Authority LinkedIn, executive publishing, award citations 8 Data + Knowledge Graphs Wikipedia, Wikidata, Crunchbase, ZoomInfo 9 Marketplaces + Ecosystems Clutch category pages, Agency Spotter, partner directories 10 Case Studies + Proof Published case studies, award wins, client outcomes 11 Community + Q&A Reddit, Quora, Lin

If you've shipped software in the last three years, you've probably watched your job description...