🦄 Lovable becomes Unicorn with $200M Series A in record time 🦄
Swedish AI coding startup reaches $1.8B valuation 8 months after launch

Swedish AI coding startup Lovable has achieved unicorn status with a $200 million Series A round led by Accel at a $1.8 billion valuation, just 8 months after launch. The Stockholm-based company enables users to create websites and apps using natural language, competing with platforms like Cursor in the AI vibe coding space. Lovable has reached impressive metrics including 2.3 million active users, 180,000 paying subscribers, and $75 million in annual recurring revenue within seven months.

The platform has generated 10 million projects to date, primarily from non-technical users creating prototypes for collaboration with developers.

The company operates with a lean team of 45 employees and attracted high-profile angel investors including Klarna CEO Sebastian Siemiatkowski, Slack co-founder Stewart Butterfield, and HubSpot co-founder Dharmesh Shah. CEO Anton Osika highlighted growth from $17 million ARR and 30,000 paying customers in February to current metrics, with enterprise clients including Klarna and HubSpot.

📄 Are Small Language Models the future of Agentic AI? 📄
Academic paper argues SLMs are more suitable for specialized AI agent

A new research paper on arXiv titled "Small Language Models are the Future of Agentic AI" argues that small language models (SLMs) are more suitable and economical than large language models for agentic AI systems. The authors contend that while LLMs excel at general conversation, the rise of agentic AI applications involves specialized, repetitive tasks with little variation where SLMs are sufficiently powerful.

The paper positions that SLMs are inherently better suited for many agentic system invocations due to their operational and economic advantages. For scenarios requiring general-purpose conversational abilities, the researchers advocate for heterogeneous agentic systems that invoke multiple different models. The paper outlines a general LLM-to-SLM agent conversion algorithm and discusses potential adoption barriers.

🎵 Suno Launches v4.5+ with Vocal and Instrumental Addition 🎵
Convert complex documents to clean digital formats

AI music startup Suno has released v4.5+, an update to its "newest and most expressive" model that launched in May. The new version introduces advanced workflows that allow users to add vocals to tracks and upload voice recordings and vocal stems before incorporating Suno-generated instrumentals.

CEO Mikey Shulman described the update as "not a minor upgrade" but rather "a glimpse of the future" enabling "more people making more music, with better tools, in more intuitive ways." This release continues Suno's trend of adding features geared towards human musicians rather than just automated song generation.

The update follows Suno's recent acquisition of browser-based DAW WavTool and the appointment of music industry veteran Paul Sinclair as Chief Music Officer. However, the company continues to face copyright infringement lawsuits from major labels and German collecting society GEMA as it expands its capabilities in AI-generated music production.

INTERESTING TO KNOW

🎬Firefly adds Video Controls and Sound Generation🎬

Adobe has significantly expanded Firefly's video capabilities with improved motion fidelity, advanced video controls, and a new Generate Sound Effects feature. The updated Firefly Video Model now offers smoother transitions and lifelike accuracy with enhanced capabilities for dynamic landscapes, animal motion, atmospheric elements, and 2D/3D animation.

Adobe is integrating partner AI models including Runway's Gen-4 Video, Google Veo3 with Audio, and upcoming additions like Luma AI's Ray 2, Pika 2.2, and Moonvalley's Marey. New creative tools include Composition Reference for Video (maintaining visual flow across scenes), Style Presets (claymation, anime, line art), and Keyframe Cropping for format-specific content. The Generate Sound Effects (beta) feature creates custom audio using text prompts or voice guidance, while Text to Avatar (beta) transforms scripts into avatar-led videos.

⚠️ Anthropic restricts Claude Code usage limits ⚠️

Anthropic has quietly tightened usage limits for Claude Code without informing users, causing significant disruption for heavy users including those on the $200-a-month Max plan. Users are receiving "Claude usage limit reached" messages with only hours-long reset times, despite believing they have substantial usage allowances remaining. Many complaints have surfaced on Claude Code's GitHub page, with users reporting inaccurate usage tracking and inability to make progress on projects. The issues coincide with broader network problems at Anthropic, including API overload errors and six separate issues in four days, though the status page shows 100% uptime. Anthropic's tiered pricing system sets limits without guaranteeing absolute access levels, with the Max plan promising 20x higher limits than Pro, but no clear values provided to users.