My obituary for programming
Simon Willison’s Weblog 4 days ago 22
My obituary for programming argues that coding agents and LLMs have shifted programming work away from human career programmers who liked building and solving logic puzzles.
AI Assistants
865 summarised stories about AI Assistants, each linking back to the original source. Browse all topics →
Monday, 19 October 2026
Tolan snags $30M for its AI companion that helps users navigate life
TechCrunch 5 days ago 33
Tolan, an AI companion for daily life management and emotional support, raised new funding to expand its runway. The startup secured another $30M, doubling its total raised to date. It plans to keep growing its AI companion product as it faces competition from newer AI agents and scrutiny around AI emotional-support risks.
Replika’s founder built the AI assistant she pitched at Disrupt 12 years ago
TechCrunch 5 days ago 42
Wabi’s founder Eugenia Kuyda launched Wabi for all users after initially pitching a chatbot vision at TechCrunch Disrupt 12 years ago with her earlier startup Luka, which later became Replika. Wabi 2.0 adds a multi-threaded, multiplayer agent system that can spin up threads and apps automatically when connected to email. The product shifts from building and sharing prompt-made mini apps to letting the agent decide what apps to create, including group-chat style shared threads with scoped visibility.
Stage
Product Hunt 5 days ago 9
Stage is launching an interface design workflow where reviewers can pin comments on a prototype built in tools like Figma, Claude, Cursor, or GitHub. It uses Claude to fix those pinned comments in v2. The process then moves from design sign-off to creating a Linear or Jira ticket.
Harness²
Product Hunt 5 days ago 31 ● 2 sources
Google Research launched Harness², a plug-and-play framework for recursive LLM agents that evolve their prompts, tools, and memory for open-world tasks without ground-truth verifiers. It reported up to +11.5 points on professional benchmarks including WorkBuddy-Bench. As a result, agents can automatically refine and accumulate reusable “execution harness” components over time rather than relying on fixed prompts or verifiers.