Bolt just rolled out native mobile app support on its platform, letting creators go from a text prompt straight to the App Store without writing any code. It effectively brings vibe coding directly to mobile software development.
Anthropic just launched the Anthropic Economic Index to track AI's real-world impact on labor markets using anonymized Claude conversations. The data shows AI leans toward human augmentation over automation at 57% to 43%, with software engineering and writing seeing the heaviest use.
OpenAI just shared three key economic observations on the road to AGI, noting that intelligence costs drop 10x every 12 months while socioeconomic value grows super-exponentially. The post also highlights that AI agents will soon act as junior virtual coworkers, making human agency more vital than ever.
GitHub just introduced agent mode for Copilot in VS Code alongside the general availability of Copilot Edits. The agent can infer missing subtasks, propose terminal commands, and fix runtime errors on its own without manual copy-pasting. GitHub also added Gemini 2.0 Flash to the model picker and previewed Project Padawan, an autonomous agent built to resolve issues and open pull requests directly.
Lyft just partnered with Anthropic to deploy Claude for customer support, cutting resolution times for driver inquiries by 87%. The bot handles frequent questions before routing trickier issues to human staff. Notably, Lyft also revealed that generative AI now writes one in four lines of its engineering code.
Google just made Gemini 2.0 Flash generally available to everyone and introduced an experimental Gemini 2.0 Pro packing a massive 2M-token context window for heavy coding and reasoning tasks. The lineup also adds Flash-Lite, offering 1.5 Flash performance at cut-rate pricing. Gemini Advanced users can already pick the Pro model from the app switcher.
ByteDance just launched Trae, a new AI code editor targeting overseas developers. Built on Microsoft software, it lets programmers generate project-level code using OpenAI's GPT-4o or Anthropic's Claude 3.5 Sonnet—and all AI features are currently free. The app is live on macOS, with a Windows version on the way.
Perplexity just rolled out its Sonar API, letting developers bake real-time web search with citations straight into their apps. The service comes in two tiers—standard Sonar and 200k-token Sonar Pro—with base pricing starting at $5 per 1,000 searches. Why it matters: Perplexity is productizing its core consumer moat into an API, taking direct aim at Google and OpenAI.
DeepSeek just launched DeepSeek-R1, an open-weight reasoning model that matches OpenAI-o1's performance on math, code, and reasoning. It arrives under a permissive MIT license alongside six distilled smaller models, with chat and API access live immediately. Basically, developers can now freely commercialize and distill frontier-level reasoning outputs without restrictions.
Cognition just released Devin 1.2, giving its AI software engineer better in-context reasoning to understand repository structures and reuse existing code. Users can now assign tasks simply by sending voice notes directly in Slack. The platform still starts at $500 a month, now with usage-based overage options.
Keras creator François Chollet just teamed up with Zapier co-founder Mike Knoop to launch Ndea, a new research lab betting on program synthesis over raw model scaling. The team aims to combine deep learning with automated code generation so AI can adapt and reason like humans. Why it matters: Instead of throwing more compute at the same LLM architectures, Chollet is building an alternative path toward true intelligence.
UC Berkeley’s NovaSky team just open-sourced Sky-T1-32B-Preview, a reasoning model that rivals OpenAI's o1-preview while costing under $450 to train. Fine-tuning Qwen-2.5-32B on 17K curated reasoning samples took 19 hours on 8 H100 GPUs. Best of all, the entire pipeline, data, and weights are publicly available.
Microsoft just released the official weights for its 14B-parameter Phi-4 model on Hugging Face under a permissive MIT license. Previously restricted to the Azure AI Foundry platform, developers can now freely download, fine-tune, and deploy the high-reasoning model locally for commercial applications.
Sonus AI just launched its Sonus-1 family of language models, offering Mini, Air, and reasoning-boosted Pro tiers. The flagship Sonus-1 Pro with reasoning claims a 91.8% score on MATH-500 and 90.0% on HumanEval, taking aim at frontier proprietary models across math and code.
Hugging Face just launched the Synthetic Data Generator, a no-code tool that turns natural language descriptions into custom datasets. Powered by distilabel and free inference APIs, it crafts text classification and chat datasets in minutes. The output pushes directly to Argilla and the Hub, making dataset curation and fine-tuning painless for non-coders.
ElevenLabs just teamed up with Mistral, Perplexity, Hugging Face, and over a dozen other platforms to launch the AI Engineer Pack. Builders get $50 or more in credits across each participating tool, taking the financial sting out of testing multiple paid APIs when prototyping new projects.
OpenAI just graduated its Canvas workspace out of beta, making the collaborative interface available to all free and paid ChatGPT users. The update brings native Python code execution with live data visualizations, side-by-side diff editing, and integration with custom GPTs. It is rolling out on the web and Windows desktop apps right now.
Cognition Labs just made Devin, its autonomous AI software engineer, publicly available for $500 a month. Eight months after its viral debut, users can now deploy Devin in a sandboxed computing environment with Slack and IDE integrations to build apps and fix bugs end-to-end. Why it matters: This is the first major commercial test for fully autonomous coding agents.
Automattic just acquired WPAI, the startup behind WordPress AI tools like CodeWP and AgentWP. The standalone apps will be sunsetted and folded directly into Automattic's core platform, with WPAI's founders joining to lead its AI roadmap. Why it matters: It marks Automattic's second acquisition in two months, signaling a major push to make AI coding and site management a native part of WordPress.
Sakana AI researchers just developed CycleQD, an evolutionary framework that merges specialized small models into multi-skilled AI swarms without expensive fine-tuning. By using model merging and SVD for crossover and mutation, it outpaced traditional fine-tuning on Llama 3-8B tests, offering a cheaper path to balanced AI agents.
