2026-05-13
By Vadym · Generated with AI, curated by me
Wednesday lands with a pricing shakeup: GitHub is moving Copilot to token-based billing on June 1, repricing how developers pay for agent workflows. Meanwhile OpenAI ships a more reliable default model, Sierra raises nearly a billion dollars for enterprise AI agents, and a Chinese humanoid robot company delivers the first credible proof of physical AI at commercial scale.
OpenAI replaced GPT-5.3 Instant with GPT-5.5 Instant as the default model for ChatGPT and the API, accessible as chat-latest. The new model produces 52.5% fewer hallucinated claims on high-stakes prompts in medicine, law, and finance compared to its predecessor. The release also ships three new Realtime voice API models: GPT-Realtime-2 (reasoning), GPT-Realtime-Translate (multilingual speech), and GPT-Realtime-Whisper (streaming transcription) — opening new surface area for voice-agent developers.
GitHub is switching all Copilot subscribers to usage-based billing on June 1, 2026. Code completions stay in the flat fee, but Copilot Chat, cloud agents, Spaces, and third-party coding agents will consume AI Credits priced at $0.01 each. Developer forums reacted sharply, with posts describing the change as paying more for identical usage. The repricing arrives as Cursor and Windsurf have emerged as flat-rate alternatives and are actively recruiting Copilot defectors.
DeepSeek is raising approximately $4 billion in its first-ever outside investment round, targeting a valuation of $45 billion — more than double the $20 billion figure circulated weeks earlier. China’s National Integrated Circuit Industry Investment Fund (the “Big Fund”) is in talks to lead, with Tencent and Alibaba also reportedly considering participation. State involvement would deepen government alignment with a lab that has already rattled Western providers on cost and capability benchmarks.
Chinese humanoid robotics company ROBOTERA closed a round exceeding $200 million led by SF Group, HSG, and IDG Capital, with Alibaba, Lenovo, Haier, and BAIC among participants. The company is running thousand-unit deliveries in Q2 2026, deployed across 10+ logistics centers with China Post and SF Group, and claims 300%+ year-over-year growth. Unlike most robotics funding announcements, this one comes with documented commercial deployments — not lab demos — making it the most concrete proof of humanoid robots at operational scale to date.
Bret Taylor’s Sierra, which builds conversational AI agents for enterprise customer service, raised $950M led by Tiger Global and GV, lifting its valuation above $15 billion from $10B in late 2025. The round reflects institutional conviction that enterprise AI orchestration is a durable category. Sierra now stands among the most highly valued pure-play AI agent infrastructure companies — a near-billion-dollar raise in a single round signals that enterprise AI is entering a capitalization phase comparable to cloud infrastructure circa 2015.
The story under today’s stories: AI is moving from adoption to infrastructure. GitHub repricing Copilot by the token isn’t a betrayal — it’s what happens when agent workflows become standard, not experimental. Sierra raising nearly a billion says enterprise orchestration is a real product category now, not a pitch deck. Ollama doubling local inference speed says you don’t need cloud for a lot of this anymore. ROBOTERA shipping thousand-unit humanoid deployments to logistics centers says physical AI commercialization has a first real data point. And DeepSeek raising at $45B with state backing says the frontier model race is getting geopolitical infrastructure under it. Every one of these moves is the same move: the industry is cementing what it built. The era of experiments is ending. The era of paying for what you use is beginning.
— Boba
Curated by Vadym