{"company":{"name":"Digitalocean","slug":"digitalocean","website":"https://digitalocean.com","category":"cloud-hosting"},"question":"What has Digitalocean shipped recently?","answer":"In the last 30 days, Digitalocean shipped 49 tracked updates. The most recent was \"17 November 2026 (microvms)\" on 2026-11-17.","window":{"days":30,"updateCount":49,"returned":49},"generatedAt":"2026-11-18T00:37:10.715Z","updates":[{"title":"17 November 2026 (microvms)","summary":"DigitalOcean's MicroVMs are now generally available in the MKC1 region. With GA, users can change a MicroVM's settings without recreating it, create access tokens scoped to a single MicroVM, add URLs for additional ports, and upload content. The content is truncated, so further GA details are not visible.","date":"2026-11-17","dateIsEstimated":false,"signalType":"product_launch","signalTypeLabel":"Launch","sourceUrl":"https://docs.digitalocean.com/notes/2026/microvms-ga/","publisher":"docs.digitalocean.com"},{"title":"DigitalOcean Droplet Snapshot Regional Copy Pricing Change Effective 1 January 2027","summary":"Starting 1 January 2027, DigitalOcean will charge the same price for Droplet snapshot copies in additional regions as the original snapshot, ending the current policy of no extra charge. Users must review and remove unused regional copies by this date to avoid unexpected billing.","date":"2026-11-12","dateIsEstimated":true,"signalType":"pricing_update","signalTypeLabel":"Pricing","sourceUrl":"https://docs.digitalocean.com/release-notes/upcoming/droplet-snapshot-regional-copy-pricing/","publisher":"docs.digitalocean.com"},{"title":"DigitalOcean Compass Council","summary":"DigitalOcean introduced the Compass Council, a program inviting current customers to participate in user research studies, quarterly sessions, and real-time feedback via Slack to influence future products and features. Participation is not guaranteed but open to any DigitalOcean customer upon request.","date":"2026-11-10","dateIsEstimated":true,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.digitalocean.com/research","publisher":"digitalocean.com"},{"title":"Now Available: Claude Fable 5.2","summary":"DigitalOcean now offers Anthropic’s Claude Fable 5.2 via Serverless Inference, Inference Router (Public Preview), model synthesis, and Evaluations. The update delivers up to 30% faster output and reduces token pricing to $8/$40 per million input/output tokens, down from $10/$50 for Fable 5.1. Eligible customers in supported regions can access it immediately.","date":"2026-11-10","dateIsEstimated":true,"signalType":"product_launch","signalTypeLabel":"Launch","sourceUrl":"https://ideas.digitalocean.com/changelog/now-available-claude-fable-52","publisher":"ideas.digitalocean.com"},{"title":"10 November 2026 (inference)","summary":"DigitalOcean deprecated several open-source models (DeepSeek V4 Flash, Llama 4 Maverick 17B, Qwen 3.5 397B, Gemma 4, gpt-oss-20b) from its Serverless Inference product as of 10 November 2026. Users relying on these models must migrate to alternatives.","date":"2026-11-10","dateIsEstimated":false,"signalType":"product_sunset","signalTypeLabel":"Sunset","sourceUrl":"https://docs.digitalocean.com/notes/2026/inference-deprecate-oss-models-nov-2026/","publisher":"docs.digitalocean.com"},{"title":"Introducing: Cuttlefish, build agents from anywhere","summary":"DigitalOcean introduced Cuttlefish, a mobile app enabling users to monitor and control agent sessions from iPhones. The app allows live tracking, approval/denial of tool calls, mid-run steering, and reviewing agent outputs across supported agents like Claude Code and CrewAI on their Managed Agents platform.","date":"2026-11-06","dateIsEstimated":true,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://ideas.digitalocean.com/changelog/introducing-cuttlefish-build-agents-from-anywhere","publisher":"ideas.digitalocean.com"},{"title":"The True Cost of LLM Inference: GPUs vs. Token Pricing","summary":"DigitalOcean published a cost framework comparing self-hosted LLM inference on DO GPU Droplets against hosted APIs, concluding that serverless open models like gpt-oss-120b are cheaper than self-hosted GPUs at most utilization levels. Dedicated GPUs only make sense for custom models, strict data control, or steady high-volume workloads above ~30% utilization.","date":"2026-11-05","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.digitalocean.com/community/tutorials/llm-inference-economics","publisher":"digitalocean.com"},{"title":"Now Available: Kimi K3.1 Pro from Moonshot AI","summary":"DigitalOcean now offers Moonshot AI's Kimi K3.1 Pro on its Inference Engine, featuring a 1M-token context window and reduced hallucination rates. The model is accessible via Serverless Inference without requiring a separate Moonshot AI account.","date":"2026-11-05","dateIsEstimated":true,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://ideas.digitalocean.com/changelog/now-available-kimi-k31-pro-from-moonshot-ai","publisher":"ideas.digitalocean.com"},{"title":"Monitor and Test AI Agents in Production using Signals and Simulations","summary":"DigitalOcean unveiled Signals and Simulations, tools to detect hidden failures in AI agents running in production. Signals monitors agent sessions at scale and surfaces issues like authentication failures or tool dependencies, while Simulations turns detected issues into repeatable tests to validate fixes before production changes. The features were demonstrated by identifying and resolving a git identity error in a coding agent that completed 40 successful runs without revealing the underlying problem.","date":"2026-11-05","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.digitalocean.com/community/tutorials/ai-agent-monitoring-testing-signals-simulations","publisher":"digitalocean.com"},{"title":"5 November 2026 (snapshots)","summary":"DigitalOcean introduced the ability to remove a Droplet snapshot’s copy from a single region while retaining the snapshot and its copies in other regions. This feature is available via the DigitalOcean API or Control Panel, allowing more granular storage management for snapshots.","date":"2026-11-05","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://docs.digitalocean.com/notes/2026/droplet-snapshot-remove-region/","publisher":"docs.digitalocean.com"},{"title":"5 November 2026 (platform)","summary":"DigitalOcean introduced a new MCP server for its documentation platform, enabling full-site search with page excerpts and product-specific filtering. The tool also supports reading documentation pages directly through the MCP interface.","date":"2026-11-05","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://docs.digitalocean.com/notes/2026/docs-mcp-search-and-page-windows/","publisher":"docs.digitalocean.com"},{"title":"Now Generally Available: Microsoft Entra ID SSO for Teams","summary":"DigitalOcean now supports Microsoft Entra ID for single sign-on (SSO) in Teams, expanding its identity provider options beyond Okta. Organizations using Entra ID can enable seamless login for their teams using existing credentials, with access managed through their directory settings.","date":"2026-11-04","dateIsEstimated":true,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://ideas.digitalocean.com/changelog/now-generally-available-microsoft-entra-id-sso-for-teams","publisher":"ideas.digitalocean.com"},{"title":"DigitalOcean Spaces Object Storage price dropped from $5.00 to $0.01 (-100%)","summary":"Product price decreased by 100%. Was $5.00, now $0.01.","date":"2026-11-04","dateIsEstimated":true,"signalType":"pricing_update","signalTypeLabel":"Pricing","sourceUrl":"https://www.digitalocean.com/products/spaces#product-20261104200430","publisher":"digitalocean.com"},{"title":"Now in Public Preview: Secrets Manager","summary":"DigitalOcean introduced Secrets Manager in public preview, enabling users to create, store, and manage application secrets (API keys, database credentials, OAuth tokens, certificates) via Cloud Console, API, or CLI. Access is controlled by DigitalOcean IAM, and secret activity is audited without exposing values.","date":"2026-11-04","dateIsEstimated":true,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://ideas.digitalocean.com/changelog/now-in-public-preview-secrets-manager","publisher":"ideas.digitalocean.com"},{"title":"4 November 2026 (inference)","summary":"DigitalOcean introduced serverless and dedicated inference support for Moonshot AI’s Kimi K3.1 Pro model on November 4, 2026. The model is now listed on DigitalOcean’s Available Models page, expanding its AI inference offerings.","date":"2026-11-04","dateIsEstimated":false,"signalType":"product_launch","signalTypeLabel":"Launch","sourceUrl":"https://docs.digitalocean.com/notes/2026/inference-kimi-k3-1-pro-model/","publisher":"docs.digitalocean.com"},{"title":"What Is DigitalOcean? Inside the AI-Native Cloud (2027)","summary":"DigitalOcean unveiled its AI-Native Cloud, integrating inference, compute, data, and agentic workloads into a single platform. The Inference Engine offers 85+ open and frontier models via an OpenAI-compatible endpoint, with an Inference Router for automatic failover and cost optimization. GPU Droplets and Bare Metal GPUs provide on-demand NVIDIA/AMD accelerators, while Managed Agents and Learning Services support production-grade agentic applications.","date":"2026-11-03","dateIsEstimated":false,"signalType":"product_launch","signalTypeLabel":"Launch","sourceUrl":"https://www.digitalocean.com/resources/articles/what-is-digitalocean","publisher":"digitalocean.com"},{"title":"Now Available: GPT-6.1 Terra from OpenAI","summary":"DigitalOcean now offers OpenAI’s GPT-6.1 Terra model through its Serverless Inference, Inference Router (Public Preview), and model synthesis services. The model delivers performance comparable to GPT-6 Astra on Agents’ Last Exam at one-tenth the cost per token and integrates seamlessly with DigitalOcean’s existing account, API, and billing infrastructure.","date":"2026-11-03","dateIsEstimated":true,"signalType":"integration","signalTypeLabel":"Integration","sourceUrl":"https://ideas.digitalocean.com/changelog/now-available-gpt-61-terra-from-openai","publisher":"ideas.digitalocean.com"},{"title":"10 Best LLM Observability Tools for AI Agents in 2027","summary":"DigitalOcean introduced three new LLM observability features—Insights, Signals, and Evaluations—integrated into its AI-native cloud platform. Insights collects metrics, logs, and traces from Droplets and GPU Droplets, Signals flags problematic agent sessions, and Evaluations scores fixes using LLM-as-a-judge metrics. All are in public preview and tied to DigitalOcean resources.","date":"2026-11-02","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.digitalocean.com/resources/articles/best-llm-observability-tools","publisher":"digitalocean.com"},{"title":"DigitalOcean Serverless Inference latency: p50, p95 and p99 TTFT for GLM-5.3, DeepSeek V3.2, Gemma 4 and Llama 4 Maverick (3,456 requests, 48 hours)","summary":"DigitalOcean measured time-to-first-token (TTFT) and tokens-per-second for its Serverless Inference across four models over 48 hours (3,456 requests). GLM-5.3, Gemma 4, and Llama 4 Maverick showed median TTFTs of ~1.0–1.1s, while DeepSeek V3.2 lagged at ~1.35s median. DeepSeek also exhibited higher tail latency (p99 up to 20.2s on long prompts). 99.74% of requests succeeded, with reasoning models like GLM-5.3 requiring up to 27s for first answer tokens.","date":"2026-11-02","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.digitalocean.com/community/tutorials/serverless-inference-latency-p50-p95-p99-ttft","publisher":"digitalocean.com"},{"title":"2 November 2026 (teams)","summary":"DigitalOcean Teams now supports Microsoft Entra ID SSO without requiring an IdP-sent team_role claim. When omitted, DigitalOcean automatically assigns a user-defined Default role during SSO configuration.","date":"2026-11-02","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://docs.digitalocean.com/notes/2026/sso-entra-default-role/","publisher":"docs.digitalocean.com"},{"title":"Request Queuing and Latency Under Burst Traffic: What Happens When Your Model Gets Overloaded?","summary":"DigitalOcean released a technical tutorial explaining how to measure and manage request queuing and latency during burst traffic for LLM inference workloads. The guide covers testing on Serverless Inference, Dedicated Inference, and self-managed setups, emphasizing metrics like goodput, P99 latency, and recovery time under 2×, 5×, and 10× traffic spikes.","date":"2026-10-30","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.digitalocean.com/community/tutorials/llm-inference-burst-traffic-queuing-latency","publisher":"digitalocean.com"},{"title":"How do I monitor and trace AI agents running in Production","summary":"DigitalOcean demonstrates how traditional logging fails to catch silent failures in AI agents, where tool calls succeed but produce wrong answers. Using DigitalOcean Managed Agents, the company shows how full-stack observability and event replays can identify and debug these issues, with tests costing ~$1.59 per session.","date":"2026-10-30","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.digitalocean.com/community/tutorials/why-agent-debugging-is-different-from-traditional-logging","publisher":"digitalocean.com"},{"title":"Now Available: Billing Insights","summary":"DigitalOcean introduced Billing Insights, a tool that provides daily spend breakdowns by model, token type, and access key, with tag, project, and product filters, CSV export, and API access. It includes a live Serverless Inference dashboard (30-day lookback, <10 min refresh) and public preview spend alerts with optional automatic pauses per model access key.","date":"2026-10-29","dateIsEstimated":true,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://ideas.digitalocean.com/changelog/now-available-billing-insights","publisher":"ideas.digitalocean.com"},{"title":"Inside DigitalOcean Simulations: how user simulation, execution, and validation work","summary":"DigitalOcean introduced Simulations, a Public Preview feature in its Evaluations product, to test AI agents through multi-turn conversations rather than single prompts. Simulations records full trajectories, uses an LLM judge to score outcomes, and compares baseline vs. candidate agent performance. It runs in Harness Runtime sandboxes and charges serverless inference rates for user simulator and judge models.","date":"2026-10-29","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.digitalocean.com/blog/inside-digitalocean-simulations","publisher":"digitalocean.com"},{"title":"How to Run Untrusted LLM-Generated Code in an Isolated Environment With MicroVMs","summary":"DigitalOcean released MicroVMs in public preview on October 29, 2026, as a secure sandbox for running untrusted LLM-generated code. Built on Firecracker, each MicroVM runs in its own isolated VM with a dedicated kernel, pauses when idle, and supports checkpoints. The feature is available in one region (MKC1) and integrates with doctl v1.173.0+ for management.","date":"2026-10-29","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.digitalocean.com/community/tutorials/untrusted-code-in-microvm","publisher":"digitalocean.com"},{"title":"Grok Bot vs Dots vs Muse: Which Agent Fits Developers in 2027?","summary":"DigitalOcean published a detailed comparison of Grok Bot (xAI), Dots (OpenAI), and Muse (Meta) as always-on AI agents, emphasizing their lack of documented APIs for programmatic control. The article positions DigitalOcean’s Hermes Agent on Managed Agents as a developer-controlled alternative, highlighting its open-source foundation and managed runtime.","date":"2026-10-28","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.digitalocean.com/resources/articles/grok-bot-vs-dots-vs-muse","publisher":"digitalocean.com"},{"title":"DigitalOcean Signals: Observability for AI Agents and Inference","summary":"DigitalOcean launched Signals, a Public Preview feature that analyzes production AI agent conversations to detect behavioral signals like user frustration, looping, tool failures, or rate limits. Signals converts raw agent interactions into structured reports with exact failure points, enabling faster triage and dataset exports for evaluation, fine-tuning, or reinforcement learning. The tool is off by default, respects privacy, and integrates with Managed Agents and Inference Engine without additional instrumentation.","date":"2026-10-28","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.digitalocean.com/blog/digitalocean-signals-observability-ai-agents","publisher":"digitalocean.com"},{"title":"DigitalOcean Managed Agents latency: cold start and resume","summary":"DigitalOcean’s Managed Agents resume latency was initially misreported as 540 ms due to polling artifacts, but corrected measurements show 249 ms at the median. New sessions take ~650 ms to become ready, with the first command landing in ~0.9 seconds. Pausing sessions reduces resume time by ~400–600 ms compared to cold starts, and idle sessions cost ~$9/month while paused sessions cost ~$0.50/month.","date":"2026-10-28","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.digitalocean.com/community/tutorials/managed-agents-cold-start-resume-latency","publisher":"digitalocean.com"},{"title":"28 October 2026 (managed-agents, agent-harness-runtime)","summary":"DigitalOcean released Cuttlefish, an iOS app for its Harness Runtime, now in public preview on the App Store. Users can follow and steer sessions, approve or deny tool calls, and manage workspace files and checkpoints directly from an iPhone.","date":"2026-10-28","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://docs.digitalocean.com/notes/2026/managed-agents-cuttlefish-ios/","publisher":"docs.digitalocean.com"},{"title":"Cuttlefish by DigitalOcean","summary":"DigitalOcean released Cuttlefish, an iOS app enabling users to run and manage AI coding agents (Claude Code, Codex, OpenCode, OpenHands, etc.) directly from iPhones. The app supports team collaboration features like shared sessions, live approvals, and sandboxed environments with pricing transparency.","date":"2026-10-28","dateIsEstimated":false,"signalType":"product_launch","signalTypeLabel":"Launch","sourceUrl":"https://apps.apple.com/us/app/cuttlefish-by-digitalocean/id6818391373","publisher":"apps.apple.com"},{"title":"9 Best Langfuse Alternatives for LLM Observability in 2027","summary":"DigitalOcean introduced DigitalOcean Insights, a built-in observability tool for its AI products (Serverless Inference, Batch Inference, Managed Agents, etc.) that requires no SDK or collector setup. It uses OpenTelemetry and is available at no additional charge during public preview, though trace export is not yet supported.","date":"2026-10-27","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.digitalocean.com/resources/articles/langfuse-alternatives","publisher":"digitalocean.com"},{"title":"Agents in Production, Part 3: Let Your AI Support Agent Send Replies, Without Giving It the Credential","summary":"DigitalOcean’s AI support agent now uses Action Gateway to broker credentials, removing API keys from the agent’s environment and preventing credential leaks. The system triaged 335 tickets, with 57 replies sent autonomously without human approval, all in non-sensitive categories. Network egress is restricted via allowlists to prevent data exfiltration.","date":"2026-10-27","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.digitalocean.com/community/tutorials/ai-support-agent-send-replies-without-credentials","publisher":"digitalocean.com"},{"title":"Three layers, all config: how DigitalOcean Managed Agents enforce NVIDIA OpenShell-aligned boundaries","summary":"DigitalOcean introduced Managed Agents in public preview, enabling AI agents to perform production tasks like migrating a backend from Supabase to DigitalOcean without direct credential access. The platform enforces three security layers—credential brokering via Action Gateway, filesystem and network boundaries using NVIDIA OpenShell policy profiles, and deny-by-default network egress rules—all configured upfront.","date":"2026-10-26","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.digitalocean.com/blog/managed-agents-enforce-nvidia-openshell","publisher":"digitalocean.com"},{"title":"What Is Batch Inference? (2027)","summary":"DigitalOcean introduced batch inference on its Inference Engine, allowing users to process large AI workloads asynchronously at up to 50% lower cost than real-time rates. The feature supports OpenAI and Anthropic models, runs within a 24-hour window, and uses separate rate limits from production traffic.","date":"2026-10-26","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.digitalocean.com/resources/articles/what-is-batch-inference","publisher":"digitalocean.com"},{"title":"Cold Start Mechanics: How Warm Pools and Model Caching Actually Work","summary":"DigitalOcean’s tutorial breaks down why AI inference services experience delays (cold starts) when scaling, attributing it to model loading times on GPUs. It details how warm pools and model caching reduce these delays, with technical breakdowns of each stage’s bottlenecks and cost trade-offs.","date":"2026-10-26","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.digitalocean.com/community/tutorials/cold-start-mechanics-draft","publisher":"digitalocean.com"},{"title":"Whats is Batch Inference? (2027)","summary":"DigitalOcean introduced Batch Inference on its Inference Engine, enabling asynchronous processing of large AI workloads at up to 50% lower cost than real-time rates. The feature supports OpenAI and Anthropic models, runs within a 24-hour window, and uses separate rate limits to avoid impacting live traffic.","date":"2026-10-26","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.digitalocean.com/blog/what-is-batch-inference","publisher":"digitalocean.com"},{"title":"26 October 2026 (inference)","summary":"DigitalOcean’s Inference service will stop supporting OpenAI’s o1, o3-mini, and GPT Image 1 models effective 26 October 2026. Users are directed to migrate to GPT-5.6 Sol for o1/o3-mini and GPT Image 2 for GPT Image 1 to maintain functionality.","date":"2026-10-26","dateIsEstimated":false,"signalType":"product_sunset","signalTypeLabel":"Sunset","sourceUrl":"https://docs.digitalocean.com/notes/2026/inference-deprecate-openai-o1-o3-mini-gpt-image-1/","publisher":"docs.digitalocean.com"},{"title":"25 October 2026 (managed-agents, agent-harness-runtime)","summary":"DigitalOcean rebuilt Harness Runtime platform templates to address CVE-2026-111497, a remote code execution vulnerability in the Brotli decoder affecting Node.js. The update impacts templates like codex and claude-code, with no further details on affected versions or remediation timelines.","date":"2026-10-25","dateIsEstimated":false,"signalType":"security_incident","signalTypeLabel":"Security","sourceUrl":"https://docs.digitalocean.com/notes/2026/harness-runtime-nodejs-cve-2026-111497/","publisher":"docs.digitalocean.com"},{"title":"Now Available: Serverless Inference Prompt Caching","summary":"DigitalOcean introduced serverless inference prompt caching, reducing costs for repeated inputs by up to 90% and cutting time-to-first-token by up to 80%. The feature includes a new parameter for cache isolation and a dashboard to track performance and savings across models like Kimi K3 and GLM-5.3.","date":"2026-10-23","dateIsEstimated":true,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://ideas.digitalocean.com/changelog/now-available-serverless-inference-prompt-caching","publisher":"ideas.digitalocean.com"},{"title":"MiniMax M3.1 Flash: Redefining Agent Performance and Cost","summary":"MiniMax unveiled M3.1 Flash, a next-generation AI model optimized for agentic tasks, coding, computer use, and cybersecurity. It delivers nearly three times faster online inference and reduces usage costs by 70% compared to its predecessor, with flat pricing of $0.15 per 1M input tokens and $0.60 per 1M output tokens.","date":"2026-10-23","dateIsEstimated":true,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.minimax.io/blog/minimax-m31f","publisher":"minimax.io"},{"title":"DigitalOcean Signals: Observability for AI Agents and Inference","summary":"DigitalOcean launched Signals in Public Preview, an observability tool that analyzes production AI agent conversations to detect 25 signal types across interaction, execution, and environment layers. It flags sessions with issues like stagnation, failures, or satisfaction, enabling targeted reviews without manual transcript analysis or costly LLM grading.","date":"2026-10-23","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.digitalocean.com/blog/digitalocean-signals-observability-ai-agents-inference","publisher":"digitalocean.com"},{"title":"Prompt Caching is GA on DigitalOcean Serverless Inference: Up to 90% Off the Context Your Agents Resend","summary":"DigitalOcean made prompt caching generally available for its Serverless Inference platform, reducing costs by up to 90% for cached input tokens and cutting time to first token by up to 80%. The feature now includes cache affinity routing, a new dashboard for tracking savings, and expanded context windows up to 256K tokens.","date":"2026-10-23","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.digitalocean.com/blog/now-available-prompt-caching","publisher":"digitalocean.com"},{"title":"23 October 2026 (billing)","summary":"DigitalOcean enhanced its Cost Analysis page with new grouping and filtering options for spend tracking, including project, tag, model, token type, and model access key. These options are now available alongside existing product-based filters.","date":"2026-10-23","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://docs.digitalocean.com/notes/2026/billing-cost-analysis-filters/","publisher":"docs.digitalocean.com"},{"title":"23 October 2026 (inference)","summary":"DigitalOcean introduced MiniMax M3.1-Flash in public preview for its Inference serverless AI service. The model is now available alongside existing options, expanding DigitalOcean’s AI inference capabilities for developers.","date":"2026-10-23","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://docs.digitalocean.com/notes/2026/inference-minimax-m3-1-flash-model/","publisher":"docs.digitalocean.com"},{"title":"22 October 2026 (app-platform)","summary":"DigitalOcean updated its PHP buildpack for App Platform to version 298, adding support for PHP versions 8.5.3–8.5.12, 8.4.18–8.4.27, 8.3.31–8.3.33, 8.3.35–8.3.36, and 8.2.31–8.2.35. The update expands runtime options for developers deploying PHP applications on the platform.","date":"2026-10-22","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://docs.digitalocean.com/notes/2026/app-platform-buildpacks-2026-10-22/","publisher":"docs.digitalocean.com"},{"title":"Critical PHP RCE Vulnerabilities: Patch Now (CVE-2026-109618, CVE-2026-109619)","summary":"DigitalOcean disclosed two critical PHP remote code execution vulnerabilities (CVE-2026-109618 and CVE-2026-109619) on October 21, 2026, with public exploit code available. The flaws affect all PHP versions from 5.x onward and allow full system compromise. Patching to supported versions (8.2+) is the only mitigation, while end-of-life versions (8.1 or earlier) have no fixes.","date":"2026-10-21","dateIsEstimated":false,"signalType":"security_incident","signalTypeLabel":"Security","sourceUrl":"https://www.digitalocean.com/blog/critical-php-rce-vulnerabilities-patch-now","publisher":"digitalocean.com"},{"title":"Product Security in the Post-AI Era","summary":"DigitalOcean’s Product Security team acknowledges that AI has fundamentally disrupted traditional software development lifecycles, replacing linear phases with a chaotic, agent-driven 'Jeremy Bearimy SDLC.' While security fundamentals remain unchanged, the practice must evolve to avoid obstructing AI agents, which now handle tasks like code generation and deployment. The company is shifting toward trust-based security programs, opinionated defaults for agents, and knowledge artifacts that agents can consume autonomously to minimize manual intervention.","date":"2026-10-21","dateIsEstimated":false,"signalType":"technical","signalTypeLabel":"Technical","sourceUrl":"https://www.digitalocean.com/blog/securing-agentic-software-development","publisher":"digitalocean.com"},{"title":"Wednesday, October 21, 2026 Security Releases","summary":"Node.js released security updates on October 21, 2026, fixing 11 vulnerabilities across versions 26.x, 24.x, and 22.x, including high-severity issues in DNS resolution, TLS hostname verification, Brotli decoder, and crypto functions. Affected systems must upgrade to patched versions immediately.","date":"2026-10-21","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://nodejs.org/en/blog/vulnerability/october-2026-security-releases","publisher":"nodejs.org"},{"title":"21 October 2026 (inference)","summary":"DigitalOcean introduced the DeepSeek V4.1 Flash 1017 model to its Inference service, enabling serverless AI model deployment. The model is now available for immediate use alongside existing options.","date":"2026-10-21","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://docs.digitalocean.com/notes/2026/inference-deepseek-v4-1-flash-1017-model/","publisher":"docs.digitalocean.com"}],"attribution":{"source":"Spyingbee","url":"https://spyingbee.com/updates/digitalocean","citation":"Spyingbee, \"Digitalocean updates\", https://spyingbee.com/updates/digitalocean (retrieved 2026-11-18)"}}