Model routing
Cloudflare AI Gateway: model aliases and fallback
AI Gateway model aliases arrived on November 4. Organize models and fallback without redeploying while preserving format, permission and cost controls.
· 2 minRead guide
Codifly / Developer library
From judgment to code.
Guides to build and run products: deployments you can repeat, infrastructure you can audit, AI you can evaluate and processes that survive a retry.
63published articles
6comparison categories
75products and plans
0sign-ups required to read
Latest
Read. Try. Decide.
From the first deploy to running in production: context to understand the decision and examples to bring it into your application.
- RAG budget and freshness
Cloudflare AI Search: December billing and 24-hour sync
Cloudflare postponed AI Search billing to December 1 and changed default sync to 24 hours. Work through a cost scenario and check answer freshness.
· 2 minRead guide - Model migration
Gemini 2.5: migrate Pro and Flash before December
Google announced Gemini 2.5 changes on November 2. Review the December 1 automatic targets, parameters and costs before migrating.
· 2 minRead guide - Image generation
GPT Image 2.6: Sunburst, Flare, pricing, SVG and GIF
OpenAI released Sunburst and Flare on November 4. Compare published rates and prepare an SVG test without mixing image and text token prices.
· 2 minRead guide - Identity and agent access
Supabase MCP with Claude: SSO, permissions and offboarding
Since November 3, Claude can issue identity assertions for Supabase MCP. Review eligibility, permission scope and offboarding before enabling it.
· 2 minRead guide - Applied AI
Gemini 4 Carbon: what Google announced and what you cannot use yet
Google introduced Gemini 4 Carbon on October 28, 2026, its strongest model for cybersecurity and long tasks, but for now only for a group of defenders. What Google says, what it does not say, the price and what to do as a developer while general availability arrives.
· 5 minRead guide
By topic
The whole library
Applied AI
9 guides- Gemini 4 Carbon: what Google announced and what you cannot use yetGoogle introduced Gemini 4 Carbon on October 28, 2026, its strongest model for cybersecurity and long tasks, but for now only for a group of defenders. What Google says, what it does not say, the price and what to do as a developer while general availability arrives.
- GPT-6.1 Terra: pricing, limits and the long-prompt surchargesOpenAI launched GPT-6.1 Terra at $1 input and $5 output per million tokens. What it brings, what it does not support, how the bill changes with prompts over 272,000 tokens, caching, Batch and fast mode, and what a month costs in dollars and pesos.
- AI model retirement calendar: OpenAI, Anthropic and Google through April 2027Which models stop working between October 2026 and April 2027, with each provider's date and recommended replacement, a script that counts the days left and what changes in Claude's parameters.
- Decision models: Jev, Microsoft-Decision-1 and OpenAI's Decisions APIIn October 2026 OpenAI and Microsoft launched models that, like Jev, do not write text but return probabilities over closed options. How they compare, what they cost, how to normalize their answers and how to measure their calibration.
- Gemini 4 Argon: pricing, limits and what changes on December 1What Google published about Gemini 4 Argon for developers: a unified 1M-token window, an introductory rate that doubles on December 1, what caching covers and a monthly cost calculation in dollars and pesos.
- Muse Wonder and Muse Spark Open: what Meta's new models mean for developersMeta launched Muse Wonder, its first frontier model, and Muse Spark Open, an 816-billion-parameter model that runs on Macs. What is known, what is not, how much memory it takes and how its price compares with Gemini 4 Argon.
- TypeSafe's Jev: what a decision model is and how to use it (Noul, Choice and Score)Jev does not write text: it evaluates a state against typed questions and returns calibrated probabilities. How to call it, how to turn its answer into an action, what it costs against an LLM and when to use it.
- How to Choose an LLM for Production Without Relying on LeaderboardsChoose an LLM for production with a real eval set, cost per accepted task, p50/p95 latency, and a safe fallback for when the model fails.
- Coding agents in the enterprise: the challenge is infrastructureCoding agents already run real engineering tasks. The sandboxing, permissions, CI/CD and observability you need before letting one near production.
DevOps & platform
9 guides- Node.js and the Brotli flaw (CVE-2026-111497): how to check if your application is exposed and how to patchOn October 21, 2026 Node.js fixed an out-of-bounds write in its Brotli decoder that is already being exploited. Which versions fix it, who is really exposed, a version checker and the other nine vulnerabilities in the same release.
- PHP2Shell (CVE-2026-109618 and CVE-2026-109619): how to check if you are affected and how to patchTwo critical remote code execution flaws in PHP, fixed on October 21, 2026 in versions 8.5.12, 8.4.27, 8.3.36 and 8.2.35. What happens, who is exposed, how to check your version, what temporary mitigations exist and what to do after patching.
- October 2026 security patches: Next.js 16.4.1 and PgBouncer 1.26.1What the Next.js security updates of September 30 and October 14, 2026 and PgBouncer 1.26.1 fix, who is exposed in each case and a version checker.
- App hosting prices in 2026: Vercel, Netlify, Railway and Fly.ioThe plans and billing model of four platforms for deploying applications, read on their official pages in October 2026, with two calculated cases and the limits of the free plan.
- Cloud regions in Latin America: AWS, Google Cloud and Azure, and what each costsWhich regions AWS, Google Cloud and Azure have in Brazil, Mexico and Chile, why none is in Colombia, what S3 storage costs in each and how to measure latency from your location.
- Terraform is not a platform: the layer you're missingA map for knowing when IaC stops solving the developer experience, and how to measure the next step.
- From DevOps to MLOps and platform engineering: what to learnWhich skills to prioritize to move from DevOps into MLOps or platform engineering and stay relevant as AI automates day-to-day operations.
- IT infrastructure trends that will define 2026IT infrastructure in 2026: AI-first design, edge, multi-cloud, AIOps, SLO-driven observability, zero trust and sustainability. What changes and where to start.
- The real challenges of DevOps and how to overcome themThe 11 most common DevOps adoption challenges, from cultural resistance to cloud cost, and concrete practices to overcome each one without slowing delivery.
FinOps
8 guides- AI API prices in November 2026: rates and real costsUpdated standard rates for models from OpenAI, Anthropic, Google, Mistral, DeepSeek and other providers. Compare units, conditions and cost per task.
- Kimi K3, GLM-5.3, Qwen3.8 Max and Muse Spark pricing: an LLM Stats comparisonPer-million-token rates from Moonshot, Z.ai, Alibaba and Meta compared with other models, using updated third-party indices and a reproducible monthly example.
- What an AI API really costs in Latin American currencies (September 2026)The 17 models in Codifly's open dataset converted to Colombian, Mexican, Chilean and Argentine pesos, reais and soles, with a dated exchange rate, a monthly example and the formula to add your own taxes.
- AI API prices in September 2026: what every model costsOfficial prices for 17 models from OpenAI, Anthropic, Google, Mistral, DeepSeek and xAI, why output tokens drive the bill, and a worked monthly example.
- Object Storage Real Cost: S3 vs R2 vs B2 vs SpacesEstimate what a bucket really costs: GB-month, GiB vs GB, operation classes, egress and minimums, with a worked example across S3, R2, B2 and Spaces.
- The hidden cost of growing on AWSHow to tell healthy growth from spend that lost its owner: idle resources, commitments, environments and data transfer.
- In-house DevOps or a platform: how to think about TCOAn honest way to add up salaries, maintenance, incidents and speed without dressing up the result.
- Why SMBs Worry More About Cloud Cost Than SecurityYour cloud bill shows up every month; security risk doesn't. Why cost and security are usually one governance problem for SMBs, and the controls that fix both.
Backend & data
6 guides- How to calculate and validate the Colombian NIT check digit, with codeThe modulo-11 algorithm with prime weights used by Colombia's NIT, a worked example and Python and JavaScript functions tested with real cases, plus what the validation cannot tell you.
- Managed database prices in 2026: Neon, Supabase, PlanetScale, Turso, Cloudflare D1 and LayerbaseSix managed database providers: flat plans, usage billing, limits and three calculated workloads. Includes Layerbase, checked in November 2026.
- Supabase Free vs Pro: differences, limits and when to upgradeA point-by-point comparison of Supabase's Free and Pro plans according to its official page in October 2026: database, pausing, backups, egress and a calculated rule for deciding.
- Supabase pricing: Free, Pro and real costs in 2026Supabase Pro from $25/month per organization: multiple-project costs, Free and Pro limits, and charges the spend cap does not cover. Examples and official sources.
- Turso vs Supabase: pricing, billing model and which to chooseTurso and Supabase compared by billing model, monthly rates reviewed in November 2026 and three calculated workloads with reproducible code.
- PostgreSQL: Find the Right Index with EXPLAINRead a PostgreSQL EXPLAIN plan, design a composite index for a real query, and add keyset pagination without breaking tenant isolation.
AWS & governance
5 guides- Dev, staging and production: what each environment is and how to separate them on AWSWhat question each environment answers, how they really differ and how to separate them on AWS with distinct accounts, with a runnable rule to avoid touching production by mistake.
- Separating environments (dev, staging, production) on AWSHow to separate dev, staging and production on AWS: accounts and OUs vs VPCs, naming, per-environment config and secrets, and artifact promotion.
- Granting an external platform least-privilege access to AWSHow to give an external platform access to your AWS account: a cross-account role with an external ID, least-privilege policies and IAM Access Analyzer audits.
- What is the best organizational structure for your AWS account?How to structure AWS with Organizations: a clean management account, OUs per environment and member accounts. Step by step in the console, CLI and Terraform.
- Connecting GitHub to AWS without access keys (OIDC)Deploy from GitHub Actions to AWS with OIDC and IAM roles instead of access keys, and grant read-only repository access with deploy keys or GitHub Apps.
Deployment & CI/CD
5 guides- Environment variables and DATABASE_URL in production: a full guideWhat environment variables are, how your app reads them, the anatomy of DATABASE_URL and where to keep secrets without leaking them. Node and Python examples.
- Deploy Your App Fast Without EC2, VPC or IAM: A PaaS GuideWhen a PaaS beats hand-built EC2, VPC and IAM for shipping to production, and the compliance, network and scale signals that justify your own infrastructure.
- Why your build works locally but fails on deployYour build passes locally but fails on the server. The 10 most common causes (env vars, casing, versions, .gitignore, lockfile) and how to fix each one.
- Deploying a frontend from Git to AWSShip a frontend from your Git repository to AWS: private S3 behind CloudFront with OAC, or Amplify Hosting, with key-free GitHub Actions CI.
- Deploying a containerized HTTP service on AWSTake an HTTP service from your laptop to AWS: image in ECR, ECS on Fargate behind an ALB with health checks, rollback-safe deployments and scaling.
APIs & storage
3 guides- Colombia's TRM API: how to query the official dollar rate in Python and JavaScriptHow to query the Representative Market Rate from Colombia's open data, with no account or key, using Python and JavaScript functions, weekend handling and what to do when the API fails.
- Wompi: how to generate the integrity signature and verify webhooks, with codeHow Wompi's checkout integrity signature is computed and how to check that an event really comes from Wompi, with Python and JavaScript functions tested against the example in its documentation.
- Direct Uploads to S3: Presigned URLs Without Exposing Your FilesA secure presigned URL upload flow for S3: server-side authorization, per-tenant keys, quarantine, and object validation before you publish.
Architecture
2 guides- Infrastructure patterns that don't scale long-termSeven infrastructure patterns that work early on but create technical debt, operational risk and slower delivery as you grow, and how to fix each one in time.
- What is a load balancer? What it solves and what it doesn'tWhat a load balancer is and how it works: traffic distribution, health checks and failover, plus what it cannot fix in your app. Examples with AWS ELB.
Containers & Kubernetes
2 guides- Kubernetes Limits in Production: A Guide with Verifiable ThresholdsKubernetes limits to know before production: 110 pods per node, 5,000 nodes, etcd's 1.5 MiB request cap, 1 MiB Secrets, and how EKS, GKE and AKS differ.
- Dockerfile: consistent deployments on any cloudHow a Dockerfile packages your app with its dependencies so it deploys the same way locally, in staging and on any cloud. Best practices and ready examples.
Distributed systems
2 guides- How to Choose a Message Queue: SQS vs Cloudflare vs QStashSQS vs Cloudflare Queues vs QStash: pull or push, delivery guarantees, FIFO, retention, max message size, and how retries multiply your bill.
- Queues, Retries, and Idempotency: Stop Processing Messages TwiceBuild idempotent queue consumers that survive duplicate messages, crashes between commit and ack, and poison messages without infinite retry loops.
Networking & DNS
2 guides- Delegating a subdomain to Route 53 without moving your domainDelegate a subdomain to Amazon Route 53 with NS records at GoDaddy, Namecheap or another provider, without migrating the parent domain, and verify it with dig.
- Link a GoDaddy, Namecheap or Google Domains domain to AWS Route 53Delegate DNS for a domain registered at GoDaddy, Namecheap or Squarespace (formerly Google Domains) to AWS Route 53 without transferring it or breaking email.
Observability
2 guides- How to Reduce Your Datadog Bill with OpenTelemetryDatadog bills per host, GB, span and custom metric. Cut costs with OpenTelemetry Collector routing, sampling, cardinality control and a hybrid stack.
- Poorly defined SLIs and SLOs: why they cause false alertsWhen SLIs track CPU instead of user experience, alerts turn into noise. Learn to define SLIs, SLOs and burn-rate alerts that fire only on real user impact.
Semantic search
2 guides- Embedding prices in 2026: Cohere Embed 5, Voyage 4, Mistral and morePrice per million tokens for Cohere, Voyage AI and Mistral embedding models as read on their official pages in October 2026, the cost of storing the vectors and a worked example of 2 million chunks.
- Embeddings and RAG: Design Retrieval Before the PromptA practical RAG guide: chunking, per-tenant filters, measuring recall@k, and migrating vectors without mixing incompatible embedding models.
Identity and agent access
1 guideImage generation
1 guideInfrastructure as code
1 guideModel migration
1 guideTechnical judgment
A useful decision shows its assumptions.
Guides separate vendor documentation, illustrative examples and editorial judgment. Comparisons link their sources and keep unknown values as “No data”.
See the methodologyThe Codifly briefA clearer perspective.
A clearer perspective.
In your inbox.
Analysis, guides and new technology comparisons.