标签：#p

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

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

Corro varios agentes respaldados por Bedrock en producción: análisis de documentos, emparejamiento de contenido, búsqueda en registros, búsqueda semántica. Esta es una pasada honesta sobre el OWASP Top 10 para aplicaciones LLM desde el lado de la implementación : el código de verdad que defiende cada riesgo y, igual de importante, las categorías donde mi respuesta es "parcial" o "todavía no". Primero el modelo de amenazas Un flujo de agentes es un pipeline: entrada no confiable → prompt → modelo → parseo → actuar. Cada flecha es una superficie de ataque. Antes de cualquier control, la pregunta más útil que me hice fue esta: ¿qué puede HACER de verdad un agente si el modelo está completamente manipulado? Mi respuesta dio forma a todo lo de abajo. Los agentes son mayormente-de-lectura : llaman a un modelo, leen filas acotadas de la base de datos, y escriben resultados de análisis con llave del usuario que pide. Sin shell, sin SQL arbitrario, sin llamada a herramientas. El radio de impacto es chico por construcción, que es el control más barato que existe. TL;DR, estatus honesto OWASP LLM Mi estatus El control LLM10 Consumo sin límite Fuerte Límite de tasa + cortacircuitos de costo mensual + topes de tokens por modelo LLM06 Agencia excesiva Fuerte (por diseño) Sin llamada a herramientas; mayormente-de-lectura; escrituras acotadas a quien llama LLM01 Inyección de prompt Parcial Contenido del usuario enmarcado como DATOS (delimitadores + preámbulo anti-inyección) LLM02 Divulgación de info sensible Parcial Limpieza de PII por regex antes del modelo; exclusiones auditadas LLM05 Manejo inadecuado de salida Parcial Validación de esquema + chequeos de fundamentación + sanear-antes-de-renderizar LLM07 Fuga del system prompt Parcial Registro versionado de prompts + regla anti-eco LLM08 Vector/Embedding N/A (todavía no construido) (nada) AuthN/Z + interruptor de apagado Fuerte Llave interna, gateo por cuota/gama, deshabilitado por agente LLM10 Consumo sin límite: empieza aquí, e

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.

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

Every developer I know has the same problem: knowledge scattered across five places at once. Browser bookmarks they never re-read. Notion docs that become graveyards. Slack threads with critical context that disappear into the archive. README files that contradict each other. Stack Overflow answers bookmarked with zero recall of why. I tried most of the "second brain" setups and none of them stuck until I figured out why they kept failing: generic productivity systems are not built for how developers actually think and work. A developer's knowledge is fundamentally different from a writer's or a manager's. It is: Code-linked (a note about a library is useless without the actual code it explains) Decision-heavy (architecture decisions need context, rationale, and alternatives considered) Debugging-intensive (solutions to bugs need the exact error message, environment, and what you tried) Time-sensitive (that API migration note is only relevant for a 3-month window) Here is the structure that actually worked. The Core Structure 00-Inbox/ 10-Projects/ 20-Areas/ - Language: Python/ - Stack: AWS/ - Domain: Auth/ 30-Resources/ - Libraries/ - Tools/ - Patterns/ 40-Archive/ The key insight: Resources are evergreen, Projects are temporary, Areas are ongoing responsibilities. A note about how JWT works lives in 30-Resources/Domain-Auth/ . A note about implementing JWT for the current sprint lives in 10-Projects/Sprint-42-Auth-Revamp/ . When the sprint is done, the project gets archived. The JWT fundamentals note stays forever. The Templates That Made It Click Architecture Decision Record (ADR) # ADR-042: Use Postgres over DynamoDB for user sessions Status: Accepted | Date: 2026-06-22 ## Context We need session storage that supports complex queries for the audit log feature. ## Decision Postgres with connection pooling via PgBouncer. ## Alternatives Considered - DynamoDB: rejected (query limitations for audit log requirements) - Redis: rejected (not durable enough for complian

1. Introduction Django's ORM is one of its greatest strengths. It abstracts away raw SQL, lets you express database operations in clean Python, and gets you productive fast. But that convenience comes with a hidden cost: if you're not deliberate about how you fetch related objects, you'll silently generate far more queries than you intend — and you won't notice until your app slows to a crawl in production. The most common culprit is the N+1 query problem : a pattern where fetching a list of N objects triggers an additional query for each one, resulting in N+1 total round-trips to the database. At ten rows it's invisible. At ten thousand rows, it's a disaster. Django provides two tools to fix this: select_related and prefetch_related . This article explains how each one works internally, when to use which, and how to combine them effectively — with before/after examples and real query counts throughout. 2. Understanding the N+1 Problem Consider a simple blog with posts and authors. You want to render a list of posts, showing each post's title and its author's name. Models: # models.py from django.db import models class Author ( models . Model ): name : str = models . CharField ( max_length = 100 ) class Post ( models . Model ): title : " str = models.CharField(max_length=200) " author : Author = models . ForeignKey ( Author , on_delete = models . CASCADE , related_name = " posts " , ) The naive approach: # views.py from django.db import connection from .models import Post def list_posts () -> None : posts = Post . objects . all () # Query 1: fetch all posts for post in posts : print ( f " { post . title } by { post . author . name } " ) # ^^^ Query 2, 3, 4, ... N+1: one per post For 100 posts, this produces 101 queries . Django lazily fetches post.author the first time you access it on each object. Each access hits the database separately. You can verify this with django.db.connection.queries (requires DEBUG = True ): from django.db import connection , reset_queries

Learn how Jason Liu uses Codex to preserve context, manage complex projects, and help work continue beyond a single prompt.

Samsung Electronics deploys ChatGPT Enterprise and Codex to employees worldwide, marking one of OpenAI’s largest enterprise AI rollouts.

У большинства self-hosted VPN-панелей одна и та же боль: Docker-стек, внешняя БД, реверс-прокси и куча конфигов, которые надо связать между собой, прежде чем хоть что-то заработает. Мне хотелось наоборот — что-то, что можно закинуть на свежий VPS и поднять меньше чем за минуту. Так появилась РосПанель : self-hosted панель для администрирования личного VPN-сервера на Xray-core , который поставляется одним статическим бинарником . React-фронтенд вшит через go:embed , база — встроенный SQLite, отдельного веб-сервера нет. Поставил, открыл, добавил юзера. Главная идея: один бинарник, ничего лишнего Цель, которая определила всё остальное, — радикальная простота. В отличие от Marzban и 3x-ui, у РосПанели нет Docker-обвязки, нет внешней БД и нет отдельного веб-сервера, который надо настраивать. Всё живёт в одном исполняемом файле: Веб-интерфейс собирается в web/dist и вшивается в Go-бинарник на этапе сборки. Состояние хранится в SQLite (чистый Go-драйвер modernc , то есть без CGO ). Конфиг Xray всегда генерируется из базы и применяется супервизором — SQLite это единственный источник правды, а не JSON, который правят руками. В итоге деплой — это просто положить бинарник и systemd-юнит. Никакой оркестрации, ничего не надо держать в синхроне. Что она на самом деле делает РосПанель — это панель управления (control plane), а не VPN-клиент. Она настраивает и обслуживает ваш собственный сервер: генерирует конфиг Xray, выдаёт ссылки на подписки и показывает статистику. Протоколы из коробки — один конфиг Xray, один набор учёток: VLESS-Vision (TCP/443 + uTLS-fingerprint) Trojan-WS (через fallback на 443) Hysteria2 (UDP с port-hopping) VLESS-gRPC-REALITY (отдельный порт, маскировка под чужой TLS) Маскировка — панель спрятана за секретным путём. Любой другой путь отдаёт сайт-заглушку (11 готовых шаблонов), так что сервер неотличим от обычного хостинга. Без знания /<secret>/ форму логина не найти. TLS, который просто работает — ACME через Let's Encrypt или ZeroSSL, авто-продление и self

Modern cloud-native developer environments are fantastic. Frameworks like .NET Aspire have revolutionized local development by providing a unified developer dashboard, automatic service discovery, and OTLP telemetry collection. But what happens when it's time to deploy your microservice topology on-premises? If your target is IIS on Windows Server or a systemd service on Linux , you've likely realized that deploying the standard .NET Aspire stack on-prem is a complex puzzle. There is no native hosting model for IIS, gRPC telemetry port mapping is fragile, and the dashboard's constant WebSocket connections can consume excessive resources on on-premises virtual machines. Enter AppPipe.Hosting —a lightweight, developer-friendly NuGet package designed specifically to bring the best features of .NET Aspire to your on-premises environments. The Problem: On-Premises Microservice Orchestration is Hard While cloud platforms have native orchestration (like Kubernetes or ECS), traditional Windows and Linux environments still host a massive volume of enterprise applications. When running microservices on IIS or Linux servers, developers face three major friction points: Port Conflicts & Service Discovery : Dynamically assigning ports to multiple microservices in IIS or systemd and injecting them into dependent services is tedious. Telemetry Aggregation : Running an OpenTelemetry Collector just to aggregate traces, logs, and metrics for a small on-prem cluster is heavy and complex to configure. Resource Exhaustion : Standard Blazor Interactive Server dashboards maintain constant WebSockets and high memory usage, which quickly drains limited hosting environments. The Solution: AppPipe AppPipe is a lightweight alternative that integrates a routing gateway, an in-memory telemetry store, and a visual dashboard directly into a single library. graph TD Client(Browser/Client) -->|HTTP| Gateway subgraph User's Application Space Backend1[Backend Microservice A] Backend2[Backend Microserv

Hey DEV community! 👋 How many times a day do you need to format a messy JSON string, convert a CSV file, or parse a timestamp? And how many times do you find yourself pasting that data—which might contain API keys, user emails, or proprietary code—into a random website you found on Google? We’ve all done it, but in an era of constant data leaks, it’s a massive security risk. That’s exactly why I built Parsify . What is Parsify? Parsify is an all-in-one data converter and developer toolset designed to handle your daily formatting, parsing, and data manipulation tasks completely offline and client-side. No servers, no tracking, and absolutely no data leaks. Everything happens right inside your browser sandbox. 🚀 Key Features 100% Secure & Offline: Your data never leaves your local machine. Once the page loads, you can literally pull your internet plug and it will still work perfectly. All-in-One Toolkit: No more bookmarking ten different sites for ten different tasks. From JSON formatting and base64 encoding to data conversions, it’s all under one roof. Built for Speed: A clean, lightning-fast UI with batch-processing support to keep your workflow uninterrupted. Privacy by Design: Zero tracking scripts, zero ads, and zero database logging. Why I Built It Most online utilities are bloated with tracking pixels, pop-up ads, and cookies. Worse, you have no idea what happens to the data you paste into their input fields. As a developer, I wanted a tool that felt like a local desktop app but possessed the accessibility of a web app. Parsify is the bridge. It gives you the convenience of a web utility with the strict security boundaries of local execution. Check It Out (It's Free!) If you’re tired of compromising on data privacy for quick utilities, give it a spin: 👉 parsify.tools I’m actively working on adding more tools and converters to the suite. I would absolutely love to get the DEV community's feedback! What converters or formatting utilities do you use daily that you

Contacts are messier than they look. A user's real address book is spread across the people they've saved by hand, the people they've emailed often enough that the provider auto-collected them, and the colleagues in their company directory. Google exposes these through the People API; Microsoft through Graph; both model the data differently and split it across sources you have to query separately. The Nylas Contacts API unifies all of that behind one schema and one grant_id . You read saved contacts, auto-collected contacts, and directory contacts through the same endpoint, create and update entries that sync back to the provider, and organize them into groups. This post walks the contact surface from the HTTP API and the Nylas CLI , which mirrors every operation for terminal use. I work on the CLI, so the terminal commands below are the ones I run when I'm exploring an address book. The contact model and its three sources A contact in Nylas carries the fields you'd expect — given_name , surname , emails , phone_numbers , company_name , job_title , notes — plus richer ones like im_addresses , physical_addresses , and web_pages . The schema is the same across providers, so a Google contact and a Microsoft contact deserialize into one struct. The detail that trips people up is source . Every contact has one of three sources, and they mean very different things: address_book — contacts the user saved deliberately. This is the real address book. inbox — contacts the provider auto-collected because the user emailed them. These were never explicitly saved. domain — contacts from the organization's directory (coworkers). Knowing the source matters because "all contacts" usually isn't what you want. If you're building a contact picker, the inbox source can flood it with one-off recipients the user doesn't think of as contacts. Filter by source deliberately. See the Contacts API overview for the full data model. Before you begin You need a Nylas API key and a connected accou

Meeting notes are the feature everyone wants and nobody wants to build. The hard part isn't the summary — an LLM handles that. The hard part is getting into the meeting: a bot that joins Zoom, Google Meet, and Microsoft Teams, survives each platform's waiting room and admission flow, records cleanly, and produces a transcript you can feed downstream. Each provider has its own join mechanics, and none of them ships a tidy "record this meeting" API. The Nylas Notetaker API is that bot as a service. You point it at a meeting link, it joins on schedule, records, and generates a transcript, and you fetch the recording and transcript through one endpoint. This post walks the Notetaker surface from the HTTP API and the Nylas CLI , which mirrors the whole lifecycle for terminal use and quick testing. I work on the CLI, so the terminal commands below are exactly what I run when I'm testing a notetaker against a live meeting. Two ways to run a notetaker: grant-scoped or standalone Before any code, there's one architectural choice worth understanding, because it changes the endpoint you call. A grant-scoped notetaker is tied to a connected account and lives under /v3/grants/{grant_id}/notetakers . Use it when the bot acts on behalf of a specific user — it can read that user's calendar and join their meetings as them. A standalone notetaker has no grant at all and lives under /v3/notetakers . You hand it a raw meeting link and it joins, no connected account required. This is the one to reach for when you just have a URL and want a recording — a public webinar, a meeting on an account you haven't connected, or a system that deals in links rather than users. Same request body, same lifecycle, same media output; the only difference is whether there's a grant_id in the path. See the Notetaker overview for how both models fit together. Before you begin You need a Nylas API key. If you're using a grant-scoped notetaker you also need a connected account; for standalone, the API key al

If your integration polls Nylas every minute to check for new email, you're doing too much work and still getting stale data. Polling is a tax: you burn rate limit on requests that mostly return nothing, and a message that arrives at 12:00:05 doesn't reach your app until the next poll. Webhooks flip that around. Nylas pushes a notification to your endpoint the moment something happens — a message arrives, an event changes, a contact is created — and your app reacts in real time. This post walks the webhook surface from both sides: the HTTP API that registers and manages webhooks, and the Nylas CLI , which has genuinely useful tooling for the part everyone gets stuck on — verifying signatures and testing webhooks against local code. I work on the CLI, so the terminal commands below are the ones I run when I'm wiring up a webhook receiver. Triggers and destinations A webhook has two halves: the trigger types it listens for and the destination URL it pushes to. Trigger types are dotted event names like message.created , event.updated , and contact.created , grouped into categories — grant, message, thread, event, contact, calendar, folder, and notetaker. You subscribe one destination to as many triggers as you want. The CLI lists every available trigger so you don't have to guess the names: # All trigger types nylas webhook triggers # Only message-related triggers nylas webhook triggers --category message Webhooks are application-scoped, not grant-scoped: one webhook registered on your application receives notifications for every connected account, identified by the grant_id in each payload. See the notifications overview for the full event model. Before you begin You need a Nylas API key — webhook management is admin-level, so it uses the application's API key rather than a grant. You also need an HTTPS endpoint reachable from the public internet to receive the notifications. The CLI gets the key set up: nylas init # create an account, generate an API key For local de

Scheduling features look simple until you build them. Google Calendar speaks its own REST API with events.insert ; Microsoft 365 wants Graph and POST /me/calendar/events ; Apple and a long tail of providers expect CalDAV. The moment your app needs to read a user's events, drop a meeting on their calendar, or check whether three people are free at 2pm, you're staring down three integrations that disagree on field names, time formats, and recurrence rules. The Nylas Calendar API gives you one interface over all of them. Connect a user's account once, get a grant_id , and read calendars, manage events, send RSVPs, and compute free/busy with the same request shape whether the backing provider is Google or Microsoft. This post walks the calendar surface from both sides: the HTTP API your backend calls, and the Nylas CLI for testing the same operations in a terminal. I work on the CLI, so the terminal snippets below are the commands I actually run when I'm poking at a calendar. Calendars, events, and the calendar_id A connected account has one or more calendars , and every event belongs to exactly one of them. Most operations take a calendar_id , and the special value primary resolves to the account's default calendar — so you don't need to look up an ID to act on the main calendar. One exception: iCloud doesn't support primary , so for iCloud accounts you pass a real calendar ID from nylas calendar list . An event carries a title , a when object holding its start and end times, a list of participants , an optional location , and flags like busy . That schema is identical across providers, which is the whole point: you read a Google event and a Microsoft event into the same struct. See the Calendar API overview for how calendars, events, and availability fit together. Before you begin You need a Nylas API key and a connected account with calendar scopes. The CLI gets you there in two commands: nylas init # create an account, generate an API key nylas auth login # connect

Over the course of a few posts, I will be constructing an NES emulator that will allow you to play NES ROMs. Today’s post is the first in this series, where I will be walking through the process of emulating the CPU the NES used, the MOS 6502. STRUCTURE The 6502 has 6 primary registers we need to emulate in our program. The A register is the accumulator, which is 1 byte long The X register is an index register, which is also 1 byte long The Y register is another index register, also 1 byte long The Program Counter is 2 bytes long, allowing for accessing 65536 memory locations The Stack Pointer is 1 byte long, allowing a 256 byte stack The P register is a status register that is 1 byte long, containing the CPU status flags Below is the struct I used for the 6502 typedef struct { uint8_t A ; uint8_t X ; uint8_t Y ; uint16_t pc ; uint8_t S ; uint8_t P ; } cpu_6502 ; MEMORY The 6502 can read up to 64kb of memory thanks to the 2 byte long program counter. For my implementation, the memory is just a simple uint8_t array declared like so: uint8_t memory[65536] = {0}; This just implements an array of 65536 bytes that we will load programs into, that the CPU will then step through. However, we need to have a function that is able to decode the commands. OPCODES An opcode (operation code) is a number that represents an instruction for the CPU. The 6502 has 56 distinct (useful) opcodes I will be emulating today. A full list of the opcodes can be found in this handy little table. Since there are so many opcodes, I will not go through them all. The full code can be found in my GitHub repo here: https://github.com/benjaminavachon/6502-emulator Here we can discuss a few of the operations so you can get a feel for what they do. The first one we will discuss is very simple. It is LDA; it simply loads the piece of data after the opcode stored in memory into the A register and increments the program counter to the next memory address. A simple example of this would be: LDA 0x05 This i

Half a year ago I asked a simple question: during an online call, could a short, to-the-point hint appear on my screen in a second or two — while the other person is still talking? Not an after-the-fact transcript, but help in the moment. The result is a desktop assistant (macOS + Windows). Below is an honest breakdown of what turned out to be hard, and which solutions worked. Engineering only, no marketing. Architecture in one paragraph On the device there are only two things: audio capture and a thin UI overlay. All the "brains" (provider keys, prompts, model selection) live on the server. The client gets a short-lived per-session token and streams audio; the server returns the transcript and the generated answer. I picked this split not for "security theater" but because otherwise keys and prompts would have to be baked into the binary — and both leak instantly. Hard part #1: system audio, not the microphone The mic only captures you. You need the other party's audio — i.e. the system output. And that's where the platform pain starts: macOS. For a long time there was no native "give me system audio" API; the classic path was a virtual audio device (BlackHole/Soundflower-style) or, in recent versions, ScreenCaptureKit, which can hand you a process's audio. ScreenCaptureKit turned out to be the best option: no kernel extensions for the user to install. Windows. WASAPI loopback saves you — you can grab whatever is going to the output device, without virtual cables. Takeaway: "system audio capture" is not one feature but two different subsystems for two OSes, and most of the early bugs were about permissions and device selection, not about audio itself. Hard part #2: latency is everything A hint that arrives 6 seconds late is useless — the conversation has already moved on. The latency budget has three parts: STT (speech → text). Streaming only. Batch "recognize after the phrase ends" immediately adds 1–2 seconds. The key metrics weren't "overall accuracy on a benchm