Vercel
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 11
- Jev becomes Vercel's fastest-adopted AI model, used by 35% of paid teams
TechCrunch reports that Vercel data shows Jev, the decision model from TypeSafe AI, was used by 35% of paid teams within two weeks of its limited September 15 release — nearly double the reach of the second most popular model, Sonnet 5.5 — which Vercel says is the fastest two-week adoption any model has seen. Jev is a lightweight tool that picks among a few predetermined outputs, making it cheaper and faster than alternatives and suited to classifying large volumes of data; Vercel itself adopted it to replace GPT-6 Luna for classifying harmful commands. The report cautions the figures come only from Vercel's AI gateway and may undercount major labs offering their own APIs, and notes open-weight models now account for over 70% of tokens processed there but about 25% of spend.
TechCrunch AI · 🔥 8 - Liquid AI launches d1-ultra decision model, claims it beats Jev 1.14
Liquid AI released d1-ultra on October 15, calling it the world's best decision model and saying it outperforms its own d1 and TypeSafe AI's Jev 1.14 across 112 evaluations. The company also claims d1-ultra leads all other decision models by at least 8 points on the Artificial Analysis Decision Index, and says it was specifically trained for better-calibrated, less overconfident probabilities (all performance claims are the vendor's own). The model takes text, images or both and returns probabilities in a single forward pass without generating tokens, answering yes/no (noul), multi-label choice and score questions. d1-ultra is available today on the Liquid API, billed on input tokens only at $0.042 per million input tokens, with images counted as input tokens; Liquid AI says it will soon reach Vercel and OpenRouter, and OpenRouter has already announced availability.
Liquid AI Blog · 🔥 21 - Cursor investigated whether confidential test data reached SpaceX AI's Grok Bot training
Citing internal communications and two people familiar with the project, The Information reported that in late July, Cursor engineers raised concerns that data generated by internal testing and everyday employee use of the agent Grok Bot (internally codenamed Sand) may have contained confidential information about customers, suppliers and business partners, and may have been sent to SpaceX's AI unit, referred to in the coverage as SpaceX SI. Cursor opened an internal investigation whose findings remain undisclosed, and it has since required employees to log in only with corporate email and decided that employee usage data will no longer be included in future model training; one of its lawyers said internally that the company may not have had the legal right to train on that data. The report also describes a separate summer incident in which customer information — including data tied to Verizon, Bosch and Coupa — was exposed on Cursor sales and marketing sites built on Vercel because passwords were written directly into the underlying code, prompting engineers to take the sites offline. SpaceX SI did not respond to a request for comment.
The Information · 🔥 14 - Vercel Sandbox flaw lets rogue AI agents bypass domain allowlists to exfiltrate data
Security firm AIR Security has disclosed a flaw it calls VerCellBreak in Vercel Sandbox, the isolated code-execution environment that runs AI agents in Firecracker microVMs and uses a domain allowlist to restrict which external servers the sandbox can reach. According to the researchers, an attacker who controls an AI agent inside the sandbox and can access sensitive data could use a crafted mutual-TLS client certificate whose Authority Information Access (AIA) field points to an attacker-controlled server; when the agent connects to an allowlisted server that validates the certificate and fetches that URL, the data can leave, and because the request is made by a server outside the sandbox, Firecracker isolation does not need to be broken. AIR says it reported the issue to Vercel, which per AIR co-founder and CTO Niv Hoffman still needs time to work out a fix, and that it has found no evidence of exploitation in the wild; the company also advises enterprises to control what API keys and other sensitive data agents can reach.
iThome 台湾 · 🔥 3 - Forgepoint raises $9.8M seed to automate manufacturing quotes with AI agent Frank
Manufacturing-quoting automation startup Forgepoint announced on October 14 a $9.8 million seed round co-led by Primary Venture Partners and a16z Speedrun, Andreessen Horowitz's accelerator, with angel investors from OpenAI, Brex, Vercel and SoFi participating. Its agent, Frank, acts as an "inside sales rep": it plugs into a manufacturer's existing systems and pricing rules, reads incoming RFQs and attachments, calculates a quote from the company's cost models, emails follow-ups when details are missing and escalates technical questions to human engineers when needed. Forgepoint says early adopters saw an average 30% increase in monthly quote throughput and a 55% reduction in time to produce a quote, and it cites a 2025 Gardner Intelligence study finding that more than 50% of manufacturers expect a vendor response within 24 hours.
SiliconANGLE AI · 🔥 0 - Theory 2026 AI survey: security tops pain points, Cloudflare leads agent sandboxes
Theory published the second annual edition of its AI in Practice Survey, fielded to 455 senior technical builders on 22 September 2026. Security and governance rose from sixth to first among pain points in twelve months: 43% rate it the most important pain point, with evals second at 22%; only 10% say security tooling delivers the most value today, but 18% expect it to lead in the future, the largest expected gain of any area. Routing is now standard practice, with 45% of teams routing queries across models — 23.7% on a router they trained and 21.5% on a proprietary one — while 54.7% still send every query to a single model. Among the 259 teams whose agents write and execute code, 62% host their sandboxes on Cloudflare, over three times the closest option (Other and Vercel at 17% each), followed by Modal 16%, E2B 12%, Daytona 9% and Sail 7%. Agent performance analysis is led by business intelligence tools at 56%, with coding agents at 48%, dedicated observability tools at 41% and spreadsheets at 30%.
Tomasz Tunguz · 🔥 0 - Origin App Marketplace opens to public submissions
Origin's app marketplace now integrates more than 20 apps across deployments, CI, code review, security, monitoring and project management, with partners including Cloudflare, Linear, Sentry and Socket Security. Starting today the marketplace is also open to public submissions, so any team building developer tools can publish an app for Origin. Developers build against the Origin API, authenticate with an app JWT and installation access tokens, receive webhook events and write back check runs, pull request comments and reviews; submissions require a slug, suggested tags and a demo video showing install, uninstall and core product flows, and apps are listed once approved.
Cursor Changelog · 🔥 0 - Enterprises Shift AI Spend to Open-Weight Models, Pressuring Closed-Model Pricing
Third-party data and company interviews cited in the report show US enterprises moving routine AI workloads off the priciest frontier models and onto cheaper open-weight alternatives. AlphaSense found mentions of "open-weight" or "open source" models on US earnings and investor calls in August–September were six times the year-earlier level; Vercel said open models handled 56% of tokens on its AI Gateway in August versus 7% in December; and Citi data cited by Reuters put OpenRouter's open-model token share at 65% in June, up from 34% in January. Examples include AT&T running roughly 40% of AI workloads on open models with plans to reach 70% within a year, Tinder and Harvey routing routine tasks to open models, and Cursor cutting the cost of building a browser from scratch from over $10,000 with GPT-5.5 alone to $1,339 using its own Composer model plus Anthropic Opus 4.8.
36氪 人工智能 · 🔥 0 - Japanese ex-AV actresses and idols are out-building big-tech AI PMs with vibe coding
A commentary published by Huxiu AI argues that Japanese former adult-film performer Noa taught herself "vibe coding" with Claude Code, shipped a polished personal site and open-sourced a structured image-prompt set on GitHub, while former Juice=Juice member Miyamoto Karin built a tour fan app in about six weeks — roughly 15 working days and 60 commits — on Claude Code with Cloudflare Workers and D1; former performer Asami Mirai pivoted to AI video and app projects. The piece contrasts them with Chinese big-tech AI product managers and cites a LatePost report that Doubao has over 200 million DAU but under 1 million yuan in daily revenue against millions in daily compute cost, along with team changes on its general chat team (Doubao's PR denied layoffs, saying 11 staff were affected, three of whom left). The author argues falling technical barriers let user-savvy individuals close the gap with big platforms.
虎嗅 AI · 🔥 0 - Polylane for Vercel pitches an AI on-call engineer for Vercel apps
A Product Hunt listing introduces Polylane for Vercel, described in one line as "An AI on-call engineer for your Vercel apps." The listing provides no further detail on pricing, features or availability.
Product Hunt · 🔥 0 - TypeSafe AI’s Jev turns frequent AI tasks into fast decisions
TypeSafe AI’s Jev is a decision model designed to return structured choices, scores, or calibrated yes/no probabilities instead of conversational text. The model launched on September 15 and drew developer attention for reported low latency and pricing, while TypeSafe AI claims of large speed and cost gains have not been independently verified.
创业邦 科技 · 🔥 0
Experience and discussion from the community
Share my Vercel experienceAsk about Vercel
Nobody has shared their experience with Vercel yet.