Linear vs Lovable

AI Research Roundup: Terminal Agents, Cloudflare Traffic, Nanochat

This newsletter edition covers recent AI research and tools. Key items include a paper on terminal agent training with self-improving tasks, a method for compressing agent screen memory, a tool for generating editable 3D scenes, and a model that predicts environment responses. Cloudflare's analysis of 206 million web sessions reveals mixed human-agent control, impacting bot detection. A new C file implementation runs a 744B parameter model efficiently, and a code graph tool supports 150+ languages. Additionally, Karpathy's nanochat project trains a GPT-2-class model for $48, and a benchmark shows Apple's SpeechAnalyzer outperforming Whisper Small. The newsletter also highlights videos on model serving and an internal agent at Linear.

• Cloudflare recorded 206 million Precursor evaluations in a day, showing sessions can shift between human and automated control.
• Colibri runs a 744B parameter MoE model from a single C file, using VRAM, RAM, and NVMe as tiers.
• Karpathy's nanochat trains a GPT-2-class model for about $48 on eight H100s.

Lovable acquires Sutro to make software easier to explain and easier to trust

Lovable has acquired Sutro. Founder Tomas Halgas and three Sutro engineers join Lovable to make how software works understandable to the people building with it.

Lovable raises $400M in Series C funding at $13.3B valuation

Lovable has raised $400 million in Series C funding at a $13.3 billion valuation, led by Menlo Ventures and the Scaleup Europe Fund, managed by EQT.

Europe's unicorn boom masks early-stage funding crisis

A new report by global VC Antler, released at its European Founder Conference in London, reveals a stark contrast in Europe's startup ecosystem: while post-2020 unicorns ('rocketships') raise record funding, the early-stage pipeline is collapsing. Key findings include a drop in Series A conversion rates (from 23.3% pre-2020 to 9.3% in 2023), significant declines in pre-seed, seed, and Series A deals since 2021, and a 42-45% reduction in active early-stage investors. The report identifies two types of rocketship unicorns—'Jets' (capital-efficient, e.g., Lovable, Legora) and 'Juggernauts' (capital-intensive deep tech, e.g., AMI Labs, Ineffable Intelligence, Fuse Energy)—and estimates that $2.74B could restore pre-2020 conversion rates, potentially yielding three additional unicorns annually.

• Antler's report analyzed 760 unicorn founders, 4,129 Series A founders, and 81,055 funding rounds in Europe since 2000.
• Only 33 rocketship unicorns (founded post-2020) exist in Europe; they reach unicorn status in 2 years on average vs. 7.2 years pre-2020.
• Series A conversion rate from Seed fell to 9.3% in 2023, down from 23.3% pre-2020.