Agentic AI momentum: OpenAI superapp agents, Airspeed $20M, Perplexity efficiency, Google Gemini Spark 24/7 agents, Poetic deterministic automation, Alsa $6.5M, TrueFoundry $19M, multimodal prompting for coding agents
Key Questions
What agentic features are coming to OpenAI's superapp?
OpenAI's superapp overhaul includes new agents, with ChatGPT Work combining ChatGPT, Codex, and GPT-5.6 capabilities. Codex 2.0 was also launched as an agentic coding tool.
Which startups raised funding for agent infrastructure?
Airspeed closed a $20M Series A, TrueFoundry raised $19M Series A, and Alsa secured $6.5M seed for agent-to-agent payments. Lyzr is raising $100M at a $500M valuation.
What new agent products launched recently?
Manta AI debuted an autonomous web app testing agent while AgentKey launched a live data marketplace. AutoBots is developing multi-LLM self-improving agents.
How are companies addressing agent reliability issues?
Bespoke Labs raised $40M to build training worlds after noting 70-95% failure rates in production. Slopfix now charges $10K per week to clean up AI-generated code.
What multimodal advancements support coding agents?
Multimodal prompting techniques are gaining traction for coding agents. ByteAsk introduced a C/C++ coding agent with self-verification capabilities.
How is Meta competing in the agent space?
Meta launched Muse Spark 1.1 with aggressive pricing and strong agentic benchmarks on JobBench and MCP Atlas. The release directly challenges OpenAI and Anthropic offerings.
What infrastructure tools help deploy agents at scale?
Hugging Face partnered with SkyPilot to reduce storage and egress fees for multi-cloud setups. Vercel positions itself as the 'AWS of agents' by decoupling models from infrastructure.
What new agent ideas are emerging from investors?
Greg Eisenberg outlined 21 startup ideas including docs as product, agent escrow, and machine courts. Sakana AI is hiring for its Fugu multi-agent system.
OpenAI superapp overhaul includes agents. Claude writes 80%+ of Anthropic code. Airspeed $20M Series A. Perplexity CEO: Value Per Watt. Microsoft launches Scout. Reid Hoffman leaves Microsoft board for Manus AI. Alsa $6.5M seed for agent-to-agent payments. TrueFoundry $19M Series A for agentic AI deployment. CircleChat launched on Product Hunt — AI agent team workspace. Alibaba bans employees from using Claude Code, highlighting geopolitical friction. Multimodal prompting for coding agents gaining traction. New product launches: Edgee Claude Code Compressor V2 (fallback models), CodeMote (iPhone CLI control), AnySearch (agent search tool), Octolens (social listening API for agents). Anthropic launches Claude Cowork on mobile/web — product expansion. Anthropic faces policy backlash over billing changes and third-party harness usage. Vercel CEO Rauch argues for decoupling models from agents, positioning Vercel as 'AWS of this generation.' Bespoke Labs raises $40M for AI agent training worlds (70-95% failure rate in production). Greg Eisenberg's thread outlines 21 concrete startup ideas for agent infrastructure — docs as product, agent escrow, machine courts. geoSurge raises $12M seed for AI visibility/GEO. New signal: Slopfix charging $10k/week to delete AI-generated code — validates growing pain of AI slop in production and creates business opportunity for experienced devs. Figma acquires Bud/Orchids team (vibe-coding). Vercel acquires Better Auth (agent infrastructure). Claude adds prompt queuing. Claude Fable 5 access extended to July 12. Hugging Face tutorial on agent distillation. Codux Rust-based control plane for AI coding CLIs. Practical tip: use smallest capable model for each task, orchestrate with SOTA for planning — challenges 'biggest is best' trend. New: Prime Intellect $130M Series A at $1B (decentralized AI, $100M ARR, Ramp customer) — reinforces enterprise agent ownership trend. Hugging Face + SkyPilot partnership to cut storage/egress fees — practical unlock for multi-cloud flexibility. GPT-5.6 usage data from Howie AI (billions of tokens, agent scheduling quality on par). Mistral AI launches 8B robotics navigation model (embodied AI). New today: Lyzr $100M at $500M (AI agent startup), ChatGPT Work launch (OpenAI agent combining ChatGPT+Codex+GPT-5.6), AI Engineer World Fair 5 startups (Reducto, Raindrop, Turbopuffer, Modal, BrowserBase), FlyDocs (AI dev tool bridging agents with Linear/Jira), Opper AI (Europe's OpenRouter), Timbal AI (agent production platform), Just Ask by SEORCE (GEO tool), GPT 5.6 sol benchmark leader. Sakana AI hiring for Fugu multi-agent system. Model per use-case cheat sheet from bindureddy. GPT-5.6 cost efficiency benchmark (10x for knowledge work). New today: Meta launches Muse Spark 1.1 with aggressive pricing ($1.25/$4.25 per M tokens) and strong agentic benchmarks (JobBench, MCP Atlas) — directly challenges OpenAI/Anthropic. ByteAsk C/C++ coding agent with self-verification. AgentKey plugin marketplace for live data. Also today: OpenAI Codex 2.0 launch (agentic coding), Manta AI autonomous testing agent, AutoBots multi-LLM orchestration.