September 2026 Summaries
9 posts from Courier
Filter
Month:
Year:
Post Summaries
Back to Blog
WhatsApp Business messaging operates through approved templates that can be sent at any time and free-form session messages that are permitted only during the 24 hours after a customer messages or calls, with each customer interaction resetting that window. Meta classifies templates as marketing, utility, or authentication, treating mixed-purpose messages as marketing; utility messages should therefore remain limited to transactional updates, while authentication templates are restricted to one-time codes. Templates can include media, buttons, carousels, offers, catalogs, and Flows, while free-form messages support conversational features such as reply buttons, lists, product messages, and forms. Businesses must obtain explicit opt-in, follow messaging and content restrictions, maintain template quality to avoid pauses or disabling, and secure permission before calling customers. Marketing templates are not delivered to US phone numbers as of September 2026, making WhatsApp in the United States primarily a support and customer-conversation channel.
Sep 25, 2026
1,563 words in the original blog post.
Courier’s Device Preview feature is presented as a way for AI coding agents to verify email templates visually across real desktop, mobile, and webmail clients before publication, addressing client-specific issues such as missing buttons, broken personalization, layout failures, dark-mode readability, and incorrect responsive ordering. Using Courier Skills, an API key, realistic test data, and a vision-capable model, an agent can inspect a draft, define expected content and section relationships, run previews on Courier’s recommended client set, download screenshots, evaluate each rendering with evidence-based PASS, FAIL, REVIEW, or N/A results, and fix confirmed draft-level problems without publishing. The post recommends GPT-6 Luna at high reasoning effort based on stated visual-reasoning benchmarks and cost comparisons, while noting that Claude Sonnet 5 may be preferable for OCR-heavy reviews and that users should test models against their own labeled examples. It supplies a detailed review prompt, guidance on examining full screenshots and overlapping high-resolution tiles, examples of bugs found in actual email previews, SDK-based workflows for targeted rechecks, and implementation advice such as preserving draft state, reusing identical test data, saving temporary screenshot links promptly, and distinguishing meaningful usability defects from harmless client rendering differences.
Sep 25, 2026
3,728 words in the original blog post.
Responding to Shopify’s 2026 decision to move from React Native toward Swift and Kotlin, the post argues that this shift does not signal the end of cross-platform development but instead reflects how AI coding agents have reduced the cost of maintaining separate native implementations. It contends that React Native’s core value proposition—mapping JavaScript interfaces to native platform components to avoid duplicate work—has weakened, while Flutter remains relevant because it provides a single, fully controlled rendering engine that draws its own interface across platforms. The author emphasizes that Flutter’s approach can offer greater visual consistency and creative control, though it carries tradeoffs involving app size, accessibility, text input, new OS features, and platform-specific integrations. Shopify’s migration is presented as an unusually resource-intensive effort supported by specialized teams, automated parity checks, and custom agent tooling, making native development less practical for smaller organizations that must review and synchronize two codebases. The post recommends Flutter for new applications where custom experience and a unified rendering surface matter, native development for simple apps or products deeply tied to new operating-system features, and continued use of React Native for existing successful apps, while predicting that improving agents may make native development the default for more teams within several years.
Sep 16, 2026
2,313 words in the original blog post.
Starting October 1, 2026, WhatsApp Business Platform will charge for free-form service replies and utility templates sent within the 24-hour customer service window, ending exemptions introduced in 2024 and 2025. Both message types will be billed at each market’s utility and authentication rate, although service messages will not receive volume discounts, while marketing template pricing, inbound customer messages and calls, the 24-hour window rules, and 72-hour Click-to-WhatsApp ad entry-point windows remain unchanged. The financial effect will depend heavily on a business’s support and order-update volume and its markets; an example Brazilian retailer’s unchanged traffic raises its monthly bill 14%, while higher-rate markets such as Germany incur substantially larger increases. The change coincides with Meta’s rollout of its billable AI Business Agent Platform and may reduce the economic advantage of lengthy WhatsApp support conversations. Businesses are advised to measure service-message and in-window utility-template volumes, review thread lengths and template classifications, and assess whether some notifications are better suited to email, SMS, or in-app channels.
Sep 16, 2026
1,605 words in the original blog post.
Healthcare notification systems should separate generic alerts from protected health information (PHI), sending only minimum-necessary prompts through SMS, push notifications, and ordinary email while requiring authenticated access to view clinical details. Although in-app inboxes can hold sensitive content when protected by scoped, short-lived credentials, encryption, access controls, retention policies, audit protections, and appropriate business associate agreements (BAAs), storing PHI there also makes message bodies, metadata, and read logs regulated data; a pointer to records in the system of record can reduce this footprint. Push payloads, email bodies and attachments, lock-screen previews, carrier infrastructure, and third-party services such as analytics, crash reporting, queues, logs, and support tools can all expose PHI, so organizations should ensure each service is BAA-covered or receives no sensitive data. The guidance recommends template-level sensitivity categories and centralized routing to prevent clinical details from being sent through open channels, while noting a limited exception for patients who specifically request their own records by unencrypted email after being warned of the risk. It also highlights device protections, deliberate log retention, and the ongoing operational costs of building secure storage and compliance controls, emphasizing that notification platforms should direct users securely to the underlying health record rather than become unnecessary repositories of clinical information.
Sep 14, 2026
2,921 words in the original blog post.
BIMI, or Brand Indicators for Message Identification, enables supported email clients to display a verified brand logo beside messages, improving sender recognition and helping recipients distinguish legitimate mail from impersonation attempts. Implementing it requires first securing all sending domains and subdomains with properly aligned DKIM and enforced DMARC policies set to quarantine or reject at 100 percent, then obtaining a certificate and publishing DNS records that point to a compliant SVG Tiny P/S logo and the certificate. A Common Mark Certificate, based on 12 months of documented logo use, supports logo display in Gmail, Yahoo, and Fastmail, while a Verified Mark Certificate requires a registered design trademark but also works in Apple Mail and provides Gmail’s blue checkmark. Certificates typically cost about $1,150 to $1,500 annually, and trademark registration can extend the timeline by 8 to 18 months. Outlook does not support BIMI, and common silent failures include incomplete DMARC enforcement, incorrect subdomain configuration, invalid SVG files, certificate-logo mismatches, and DNS caching delays.
Sep 14, 2026
4,133 words in the original blog post.
Email preview tools that render messages in real email clients and return screenshots differ from browser simulations and CSS compatibility checks, with pricing generally based on one screenshot per client. The comparison reviews Courier, Litmus, Parcel, Mailosaur, SendForensics, and Unspam, finding that they vary in client coverage, dark-mode support, workflow integration, API availability, preview quotas, and whether they display the complete inbox view with sender and subject information. Courier offers 96 client configurations, editor and API integration, and plans starting at $20 monthly for 100 previews; Litmus provides the broadest stated coverage but now requires negotiated sales contracts; Parcel targets developers working in an email code editor; Mailosaur emphasizes testing emails after they have been sent; SendForensics bundles unlimited previews with deliverability tools; and Unspam offers low-cost plans but relatively high per-preview rates at smaller volumes. Several email marketing platforms, including Mailchimp, Klaviyo, HubSpot, and Constant Contact, provide embedded preview features, often powered by Litmus or other vendors, though their monthly allowances are typically limited. Key evaluation factors include whether previews use actual clients, cost at expected testing volume, distance from the authoring environment, fidelity to personalized and tracked sends, relevant client coverage, inbox-view availability, and API support for automated testing.
Sep 02, 2026
2,789 words in the original blog post.
Courier has launched the Template Metrics API, enabling users to retrieve time-series delivery analytics for individual notification templates through the same public API that powers its dashboard. The endpoint reports sends, deliveries, opens, clicks, errors, and undeliverable messages by provider and channel, with hourly, daily, weekly, or monthly aggregation and support for relative or specified date ranges. It returns complete time buckets, including empty periods, and is available in Courier’s Node, Python, Ruby, Go, Java, PHP, and C# server SDKs as well as through direct HTTP requests. Potential uses include automated deliverability reporting, delivery-rate alerts, customer-facing analytics, BI ingestion, and AI agents that investigate notification performance. Requestable periods vary by granularity and plan, with hourly data capped at seven days, daily data at 90 days, and plan-specific lookback and rate limits; all timestamps use UTC, requests cover one template at a time, and the feature is currently available in the US region.
Sep 02, 2026
1,296 words in the original blog post.
Courier has introduced Cross-Device Email Preview, a Design Studio add-on that renders email templates on 96 real clients across 21 applications, including Outlook on Windows, Apple Mail, Gmail, and major webmail services, rather than using simulated previews. Users can test unpublished drafts, review screenshots of full inbox views including sender names and subject lines, and assess client-specific behavior such as Outlook’s HTML limitations, Gmail clipping, and varying dark-mode color handling. The feature supports saved workspace-wide target sets, filtering by desktop, mobile, and web clients, as well as a recommended set of eight representative clients; preview runs generally take 10 to 60 seconds and screenshots remain available for one year. Usage is measured per client screenshot, with monthly plans starting at $20 for 100 previews, rising to $50 for 400 and $100 for 1,000.
Sep 01, 2026
1,108 words in the original blog post.