Telexed

AI news through a solo-operator lens — only what changes your day12 of 412

FILTER[All][Agents & tools][Models & API][Generative media][Infra & SaaS][ASO & growth][Indie business][Idea signals][Other][★6+ high-signal]

vercel_blog ✕clear filters

Fri, May 151 dispatches

#0412
#0412Infra & SaaS vercel_blog5 days ago
`AI Gateway` adds request-time provider ranking controls
80radar
AI GatewayAI routing service — failover across model providers
Routing can now optimize on price, first-token latency, or throughput at request time instead of Vercel's blended default. Useful when one model has many providers and the cheapest or fastest route materially changes margin or UX.
- Set sort on providerOptions.gateway to 'cost', 'ttft', or 'tps' depending on whether margin, snappiness, or long-output speed matters most.
- Ranking is computed at request time, so newly added providers, price changes, and observed latency shifts flow through without code changes.
- Fallback is strict: providers are attempted in sorted order, and the next one is used only if the higher-ranked provider is unavailable.
- sort works with Zero Data Retention filtering and with order; pinned providers stay first, then the rest follow the chosen ranking.
- Each response exposes routing metadata with a sort block showing candidates, metric values, attempt order, and health-based deprioritization for debugging.
Source: vercel.com/changelog/sort-providers-by-cost-latency-or-tRead original →
FIG-4121:1
80radar
FIG-4121:1

Thu, May 141 dispatches

#0411
#0411Infra & SaaS vercel_blog6 days ago
`Vercel` adds `Protected Source Maps` for production-only debugging
70radar
Protected Source MapsDeployment security feature — hides browser source maps behind auth
Production source maps can now stay accessible to your team while returning 404 to everyone else. It removes a common tradeoff between readable stack traces and leaking original frontend code, so existing projects should turn it on now.
- Browser .map files now sit behind Vercel Authentication, letting internal users fetch them while blocking public access with 404 responses.
- New projects get the setting enabled by default, which lowers the chance of accidentally exposing source maps on fresh launches.
- Existing projects can enable it in Settings -> Deployment Protection with no redeploy, so rollout cost is near zero.
- This keeps production debugging intact with real filenames and line numbers, without leaving unminified source structure open to anyone scraping assets.
Source: vercel.com/changelog/protected-source-maps-ship-browser-Read original →
FIG-4111:1
70radar
FIG-4111:1

Wed, May 132 dispatches

#0410
#0410Infra & SaaS vercel_blog7 days ago
`Vercel` adds `Trusted Sources` to Deployment Protection
70radar
Trusted SourcesDeployment protection feature — accepts short-lived `OIDC` tokens
Short-lived OIDC tokens can now replace shared bypass secrets for protected deployments. Safer CI and cross-project calls with less secret sprawl; worth adopting now.
- Requests now pass an OIDC token in x-vercel-trusted-oidc-idp-token; Vercel verifies signature, claims, and env match before allowing access.
- A project can access its own deployments by default, and same-team cross-project access is allowed by adding explicit source rules.
- Rules are configurable per from / environment pair, so preview, staging, and production access can be split cleanly.
- External identity providers are supported too, including GitHub Actions; one-off automation no longer needs a long-lived shared bypass secret.
Source: vercel.com/changelog/trusted-sources-for-deployment-protRead original →
FIG-4101:1
70radar
FIG-4101:1
#0409
#0409Infra & SaaS vercel_blog7 days ago
`Vercel Firewall` can generate WAF custom rules from natural language
70radar
Writing edge security rules just got much less manual. If you already ship on Vercel, this cuts setup friction enough to try now.
- The dashboard now turns plain-English intent into WAF rules, reducing the need to hand-write match logic.
- Generated rules can log, block, challenge, rate limit, or redirect traffic based on IP, path, country, user agent, and more.
- Examples are immediately practical: protect /api/webhook, block /wp-admin, or challenge non-US traffic on /checkout.
- Vercel CLI support matters because rule creation can fit existing deploy and ops workflows instead of staying dashboard-only.
Source: vercel.com/changelog/create-vercel-waf-custom-rules-usinRead original →
FIG-4091:1
70radar
FIG-4091:1

Tue, May 122 dispatches

#0408
#0408Infra & SaaS vercel_bloglast week
`AI Gateway` adds fast mode for `Claude Opus 4.7`
80radar
AI GatewayLLM gateway — unified routing across model providers
Latency drops hard without stepping down model quality: output generation is about 2.5x faster. The tradeoff is brutal at 6x standard Opus pricing, so this is for premium paths only.
- Output token generation is roughly 2.5x faster while keeping full Opus 4.7 intelligence; useful when response time is the product constraint.
- Enable it via speed: 'fast' in anthropic provider options with anthropic/claude-opus-4.7, so rollout is a config change, not a model swap.
- Pricing jumps from $5/$25 per 1M input/output tokens to $30/$150. Speed gains are real, but margins get hit immediately.
- Claude Code can use it through AI Gateway with CLAUDE_CODE_SKIP_FAST_MODE_ORG_CHECK, CLAUDE_CODE_ENABLE_OPUS_4_7_FAST_MODE, or ~/.claude/settings.json.
- Prompt caching and other standard multipliers still stack on top, so heavy agent loops can get expensive fast.
Source: vercel.com/changelog/fast-mode-for-opus-4-7-available-onRead original →
FIG-4081:1
80radar
FIG-4081:1
#0407
#0407Infra & SaaS vercel_bloglast week
Manage `Vercel Firewall` Rules from the `CLI`
70radar
Vercel FirewallSecurity feature — controls traffic rules for Vercel apps
Security rollout moves closer to deploy workflows instead of dashboard clicks. Teams using vercel can script rule changes, test safer rollouts, and hand more of this to agents now.
- The vercel firewall command now covers custom rules, IP blocks, system bypasses, attack mode, and system mitigation settings.
- This closes a real ops gap: firewall changes can live in terminal workflows alongside deploys, scripts, and incident response.
- Vercel Firewall also gets an agent-facing skill with guidance for safer rule rollout, which lowers the chance of locking out good traffic.
- Setup friction is low: update the CLI, run vercel firewall, and the feature is immediately reachable from existing Vercel workflows.
Source: vercel.com/changelog/manage-vercel-firewall-in-the-cliRead original →
FIG-4071:1
70radar
FIG-4071:1

Mon, May 111 dispatches

#0406
#0406Infra & SaaS vercel_bloglast week
`Vercel Flags` adds scheduled progressive rollouts
80radar
Vercel FlagsFeature flag service — plugs into Vercel release flow
Traffic can now ramp to a new variant by schedule instead of staying at a fixed split. That makes feature releases safer without building your own rollout logic, and it's usable now via dashboard or vercel flags rollout.
- Unlike a fixed 50/50 experiment split, each rollout stage has a target percentage plus duration, so exposure increases automatically.
- Catching regressions on a small slice first lowers blast radius before a bad release reaches everyone.
- Control is available in both the dashboard and the new vercel flags rollout CLI path, which matters for scripted releases.
Source: vercel.com/changelog/progressive-rollouts-in-vercel-flagRead original →
FIG-4061:1
80radar
FIG-4061:1

Fri, May 83 dispatches

#0405
#0405Infra & SaaS vercel_blog2 weeks ago
`Vercel` `Chat SDK` adds `Messenger` adapter
60radar
Chat SDKChat app SDK — adapter-based support for messaging channels
Meta channel support just got easier: one adapter now covers messages, reactions, media downloads, postbacks, and DMs. Useful if you want to ship customer-facing agents on Messenger without building the transport layer yourself.
- The adapter supports core Messenger flows: messages, reactions, multimedia downloads, postback buttons, and direct conversations.
- Display names are fetched automatically from user profiles, removing one small but annoying identity-mapping step.
- Vercel points to three paths: use the docs, browse supported adapters, or build a custom adapter if your channel is missing.
- This came from community work in PR #461, which suggests the adapter surface is extensible rather than locked to first-party channels.
Source: vercel.com/changelog/chat-sdk-adds-messenger-adapterRead original →
FIG-4051:1
60radar
FIG-4051:1
#0404
#0404Infra & SaaS vercel_blog2 weeks ago
`Chat SDK` adds cross-platform conversation history
70radar
Chat SDKBot-building SDK — shared conversation state across channels
Message history can now persist across channels with transcripts and identity, instead of rebuilding memory per surface. If you run support or assistant bots on web and app together, this removes a common state-sync chore and is worth adopting now.
- transcripts and identity let one user keep the same history across every platform adapter, so web, app, and other bot surfaces stop fragmenting context.
- bot.transcripts exposes 4 methods: append, list, count, and delete. The API covers write, read, usage tracking, and reset without extra glue code.
- The feature sits on top of your existing state adapter, so you do not need a separate transcript store design if you already use the SDK's persistence layer.
- list returns entries chronologically with filters, which makes replay, debugging, and lightweight conversation analytics easier to ship.
- This mainly cuts backend plumbing rather than unlocking a new UX category, but that plumbing is exactly what slows small teams down.
Source: vercel.com/changelog/chat-sdk-conversation-historyRead original →
FIG-4041:1
70radar
FIG-4041:1
#0403
#0403Infra & SaaS vercel_blog2 weeks ago
`Chat SDK` adds a web adapter for browser chat UIs
60radar
Chat SDKChat UI SDK — connects server bots to browsers
Browser chat can now connect directly with a server-defined bot and stream replies live through useChat. This cuts setup for in-product support or assistant UIs and is practical if you already ship on Vercel's AI stack.
- The new web adapter connects browser-based chat UIs to a bot defined on your server, reducing glue code between frontend and backend.
- Replies stream live to the browser through the preconfigured @ai-sdk/react useChat hook, so a basic chat surface is faster to stand up.
- The target use cases are clear: in-product assistants, support agents, and other embedded browser chat experiences.
- Vercel also points to adapter docs and custom adapter guidance, so the release is not just a demo path but an integration surface.
Source: vercel.com/changelog/chat-sdk-adds-web-adapter-supportRead original →
FIG-4031:1
60radar
FIG-4031:1

Thu, May 72 dispatches

#0402
#0402Infra & SaaS vercel_blog2 weeks ago
`Next.js` security release patches 13 flaws; upgrade to `15.5.18` or `16.2.6`
90radar
This fixes 13 issues across auth bypass, DoS, SSRF, cache poisoning, and XSS, with no reliable WAF fallback. If you run middleware.js, Server Functions, Image Optimization, or RSC caching, this is an immediate patch cycle.
- Authorization paths are exposed in multiple ways: middleware.js/proxy.js apps are affected by segment-prefetch, i18n default-locale, and dynamic route injection bypasses.
- Runtime risk is broad, not edge-case: DoS hits React Server Components, Cache Components, and the Image Optimization API, so default framework features are in scope.
- WebSocket upgrade handling has an SSRF advisory. Any app bridging internal services through upgrade requests should treat this as perimeter-breaking.
- Caching setups need scrutiny twice: one advisory targets RSC response poisoning, another targets collisions in RSC cache-busting. CDN or reverse-proxy stacks are not neutral here.
- Fixed targets are explicit: upgrade Next.js 15.x to 15.5.18, 16.x to 16.2.6, and matching react-server-dom-* packages to 19.0.6 / 19.1.7 / 19.2.6.
Source: vercel.com/changelog/next-js-may-2026-security-releaseRead original →
FIG-4021:1
90radar
FIG-4021:1
#0401
#0401Infra & SaaS vercel_blog2 weeks ago
`Vercel Flags` now supports JSON values
70radar
Vercel FlagsFeature flag service — built to fit `Vercel` deployments
Related config can now live in one flag instead of being split across several. That makes model rollouts, A/B tests, and provider failover less brittle, so it is worth using if Vercel Flags is already in the stack.
- A single JSON flag can replace separate boolean, string, and number flags tied to the same feature config.
- The example is model experimentation: one ai_model object can hold settings that previously needed ai_model, ai_temperature, and ai_max_tokens.
- This fits progressive traffic routing to a new model, controlled A/B tests, and fast provider switching during outages.
Source: vercel.com/changelog/vercel-flags-now-supports-json-valuRead original →
FIG-4011:1
70radar
FIG-4011:1

`AI Gateway` adds request-time provider ranking controls

`Vercel` adds `Protected Source Maps` for production-only debugging

`Vercel` adds `Trusted Sources` to Deployment Protection

`Vercel Firewall` can generate WAF custom rules from natural language

`AI Gateway` adds fast mode for `Claude Opus 4.7`

Manage `Vercel Firewall` Rules from the `CLI`

`Vercel Flags` adds scheduled progressive rollouts

`Vercel` `Chat SDK` adds `Messenger` adapter

`Chat SDK` adds cross-platform conversation history

`Chat SDK` adds a web adapter for browser chat UIs

`Next.js` security release patches 13 flaws; upgrade to `15.5.18` or `16.2.6`

`Vercel Flags` now supports JSON values