New this week
Releases, launches and pricing changes worth knowing about. Each one gets a one-line "who it's for" and a link to the source. Nothing here is a tested recommendation yet.
Bolt Builders Week: contest entries close 20 October, live sessions run 20–23 October
Bolt is running a contest for businesses built in Bolt, with prizes of $10,000, $5,000 and $2,500. No purchase is necessary and entrants must be 18+. Submissions close on Tuesday 20 October, and free live sessions follow from 20 to 23 October: a founder Q&A plus workshops on SEO, retention and design. Winners will be announced around 31 October. The free GLM 5.3 Flash build weekend (17–18 October) has already ended.
Who it's for: Vibe coders with a working Bolt project and a public link they can submit now.
Source →CodeRabbit renames Autofix to 'Fix comments', and Autopilot is now 'Autofix'
From 14 October, 'Fix comments' (formerly Autofix) is a one-off repair of the review comments you select, using @coderabbitai fix comments. 'Autofix' (formerly Autopilot) keeps fixing review feedback, CodeRabbit findings and failing CI on a coding-agent pull request until it's ready to merge. The old commands keep working as aliases for a limited time.
Who it's for: Teams whose scripts or habits rely on CodeRabbit's old command names.
Source →Descript's AI Speech now runs on ElevenLabs v4
Descript's 12 October release notes say AI Speech now runs on ElevenLabs v4, and that voice clones 'sound closer to the person they were cloned from'. To use it, pick v4 in your AI model settings. Generate video on paid plans now defaults to MiniMax H3 Max Turbo (free plans stay on PixVerse V4.5), and there's a new drive-wide search and 12 months of usage history.
Who it's for: Creators who patch flubbed lines in Descript with a clone of their own voice.
Source →Kling 4.0 launches with 30-second clips and up to 15 reference assets
Kuaishou says Kling 4.0 launched on 15 October. It generates up to 30 seconds natively in a single pass, handles large-scale motion and complex camera moves more steadily, and outputs up to 10-bit HDR. Omni Reference combines up to 15 references (up to 10 images, 5 videos and 7 subjects) and up to 10 keyframes. It's available through Kling's MCP and CLI, and Kling says API access for developers is coming soon.
Who it's for: Creators making AI B-roll, ads or short films who need the same character or product to look consistent across shots.
Source →Apps built in Lovable can now use GPT-Live, Claude, Gemini Omni, Lyria and Perplexity
Announced 15 October: apps you build in Lovable can use OpenAI's GPT-Live for real-time voice, the latest Claude models, Google's Gemini 4 Argon, Gemini Omni (video) and Lyria (music), plus Perplexity, Firecrawl, Parallel and Jev for cited web answers, scraping and fast routing. Describe the feature and Lovable picks a model, or name one in the chat. AI features in your app draw on your Lovable credits.
Who it's for: Builders adding a voice assistant, research panel or AI video feature to an MVP without managing API keys.
Source →Lovable can now plan your app's growth: SEO fixes, launch posts and ad campaigns
Announced 14 October: give Lovable a goal such as 'my first 100 signups' and it builds a plan from your app, audience and traffic. It can fix SEO basics, draft posts and listings for communities and directories (you publish them yourself), and set up Google and Meta ad campaigns. Nothing goes live until you approve it, ad spend is billed by Google or Meta, and Lovable says it adds no fee beyond normal credit use. Available on Free, Pro and Business.
Who it's for: Builders with a published Lovable app who don't know where their first users will come from.
Source →CodeRabbit renames its plans: Pro is now Essentials, Pro Plus is now Team
Prices stay the same according to CodeRabbit's FAQ: Essentials is $24/dev/mo billed annually ($30 monthly) and Team is $48 annually ($60 monthly). A new Advanced tier at $72/user/mo adds continuous security monitoring.
Who it's for: Teams already paying for CodeRabbit (nothing to do), and anyone comparing AI code reviewers.
Source →Cursor plans are now sold through X subscription tiers
Cursor's pricing page now lists Free, Plus ($8/mo, Grok models only), Premium ($30/mo), Super ($100/mo) and Ultra ($200/mo), bundled with Grok and X features. Third-party models are billed separately at API list price.
Who it's for: Vibe coders on Cursor: check which models your tier includes before you renew.
Source →ElevenLabs introduces Eleven v4
ElevenLabs launched Eleven v4 on 28 September, calling it 'our most emotive model yet', alongside Eleven v4 Turbo, a low-latency version for voice agents. ElevenLabs says v4 follows inline delivery tags such as [laughs] more accurately than earlier models.
Who it's for: Creators doing narration or character voices who want more expressive delivery.
Source →Lovable publishes its full Pro and Business credit tiers
Lovable's docs list Pro tiers from 100 credits for $25/mo up to 10,000 credits for $2,250/mo, with annual billing giving two months free. Students and teachers get 50% off the 100-credit Pro plan for up to 12 months.
Who it's for: Builders estimating what an MVP will really cost on Lovable.
Source →OpusClip Pro lists AI video B-roll powered by Agent Opus
OpusClip's plan comparison now includes AI video B-roll 'powered by Agent Opus' on Pro ($29/mo), alongside AI image B-roll and stock B-roll.
Who it's for: Creators clipping talking-head content who want B-roll without searching stock libraries.
Source →