September 2026 Summaries

Linq’s zero data retention feature ensures that message content, attachments, poll options, mentions, and edits are processed in memory without being stored in its database. Content is available only once: sent messages are echoed in the POST response, received messages include content in their first message.received webhook, and later webhooks or retrieval requests return empty content fields marked with a zero_retention indicator. Attachment links similarly permit one successful download before the file is deleted and subsequent requests return an error. Integrations must therefore capture and store needed information upon first receipt, resend all existing poll options when adding a new one, and explicitly provide app identity details when updating certain app cards. Unlike ephemeral messaging, which retains encrypted content for up to 24 hours to support reactions, replies, and edits, zero data retention never stores content at all. The feature is available on all plans, can be enabled per phone number, and is intended for lines handling sensitive conversations.

Caddy, a personal AI assistant created by former Loom leaders Rajiv and Connor, shifted from an installable voice-first Mac application to a text-message-based product after finding that users—including its founders—used the desktop version infrequently. Testing the same backend across email, Slack, and iMessage showed substantially stronger adoption through iMessage, which the team attributes to lower installation friction, mobile availability, and the informal, relationship-oriented nature of native messaging. Comparable findings from Lindy and VentNow suggest that text-based channels can drive more traffic, messages per session, and daily activity than standalone apps. A persistent message thread also changes agent behavior by removing fixed sessions, enabling the assistant to initiate conversations, supporting ongoing context and memory, and allowing users to send voice notes while mobile. The post presents Linq, a messaging API for iMessage, RCS, SMS, and voice, as the infrastructure used to productize Caddy.

SMS APIs enable AI agents to send and receive messages on nearly any mobile phone, but providers differ in carrier coverage, delivery reliability, latency, two-way webhook support, integration tools, uptime guarantees, and pricing. For conversational agents, SMS offers universal reach but has drawbacks including 160-character segmentation, limited rich-media and conversational features, unencrypted transmission, U.S. A2P/10DLC onboarding requirements, lower reported reply rates than iMessage, and costs that rise with message volume. Twilio, Vonage, and Plivo are presented as major SMS API options: Twilio emphasizes global reach, AI-agent tooling, conversation memory, routing, and human handoff; Vonage offers extensive carrier coverage, adaptive routing, low-code agent building, and data-redaction controls; and Plivo provides multichannel messaging, RCS capabilities, and AI workflow-building tools. The source suggests iMessage as a stronger primary channel for extended agent conversations because it supports read receipts, typing indicators, reactions, and rich media, while recommending providers such as Linq that offer iMessage with automatic RCS and SMS fallback, low latency, uptime commitments, and per-number pricing.

Vent Now, developer of Aspen, a conversational AI companion for emotional growth, tested delivering its service through iMessage rather than its branded app while keeping the backend, AI model, and testing framework unchanged. The company reported that iMessage users sent 20% more messages per session, were roughly 40% more active daily, had 40% lower churn on weekly paid plans, and showed an estimated 16% higher lifetime value, potentially because a familiar texting interface made Aspen feel more human and encouraged greater use of plans with message limits. However, the cohorts were not randomized, as iMessage access was initially introduced by subscription tier, and the findings measure engagement and upgrades among existing users rather than new-user acquisition. Vent Now is now testing text-based onboarding, proactive messages, and in-thread payments, using Linq’s messaging platform, while presenting the experiment as evidence that native messaging interfaces may substantially affect conversational AI usage.