📌 14 AI startups raise $250M+

BONUS: LLM Hallucinations Aren’t Bugs From SignalFire

October 16, 2025

TODAY’S FUNDING RECAP 🗓️

  • Liberate raises a $50M Series B — Voice AI

  • Viven raises a $35M Seed — AI digital twins

  • Dedalus Labs raises a $11M Seed — Multi-modal agents

  • Osmosis raises a $6.3M Seed — Self-learning AI agents

  • 10 more rounds caught by my funding tracker

Read time: 3 minutes

SHARE CHIEF AI OFFICE

For every 5 referrals, we’ll give you 1 month of Chief AI Office PRO for free

You currently have 0 referrals, only 5 away from adding 1 month to your PRO subscription

TOP PICKS 📢

1/4

Amount: $50M 🎉

Valuation: $300M

Round: Series B

Lead Investor: Battery Ventures

Liberate creates reasoning-based AI agents that automate core insurance workflows.

A voice assistant manages regulated conversations with policyholders at the front-end. Behind the scenes, backend agents fetch data, trigger system actions, and complete transactions in core insurance platforms. The system is trained with reinforcement learning for insurance-specific dialogue and includes a monitoring layer that flags anomalies and routes edge cases to human reviewers for compliance.

2/4

Amount: $35M 🎉

Round: Seed

Lead Investor: Khosla Ventures

Viven builds LLM agents trained on an employees internal work data.

Digital twins act as stand-ins when teammates are unavailable, returning contextual answers based on past work and communication. These agents capture departmental knowledge to have collective context. Access control is handled through a “pairwise context and privacy” model that defines permissions based on organizational relationships and role-based data visibility.

3/4

Amount: $11M 🎉

Round: Seed

Lead Investor: Kindred Ventures, Saga Ventures

Dedalus Labs offers an SDK and infrastructure stack for building production-grade AI agents.

Its platform abstracts agent orchestration, letting developers deploy non-linear, multi-tool agents in a few lines of code. Agents can switch between models, chain tools, stream outputs, and reason across providers like OpenAI, Anthropic, and Mistral. The system is built on Model Context Protocol (MCP), allowing models to invoke tools through a language-agnostic interface.

4/4

Amount: $6.3M 🎉

Round: Seed

Lead Investor: CRV, Audacious Ventures

Osmosis is a reinforcement fine-tuning platform for training task-specific AI agents.

The system enables multi-turn training with tools accessed via Model Context Protocol (MCP) to replicate production conditions. It fine-tunes agent behavior through continuous reinforcement feedback, addressing issues like brittle tool use, latency, and cascading failure in multi-step tasks. Osmosis trains models to improve autonomously across edge cases hosted on its platform or deployed externally.

OTHER ROUNDS 🤏

ESSENTIAL INTRODUCTIONS

→ Campfire raises $65M Series B — Accounting & revenue automation for startups

→ Starship raises $50M Series C — Autonomous sidewalk delivery robots

→ MatrixSpace raises $20M Series B — Portable AI radar with edge object detection and classification

→ Planyear raises $12M Seed — AI benefits consulting platform

→ Woz raises $6M Seed — AI mobile app builder

BONUS 🔥

FOR CURIOUS MINDS

LLM hallucinations aren’t bugs: The real challenges are confidence and context

AI hallucinations aren’t just bugs but signals of deeper issues in confidence and context. OpenAI and Anthropic point to LLMs that calibrate confidence while tuning imagination, accuracy, and empathy to the task.

www.signalfire.com/blog/llm-hallucinations-arent-bugs

THAT’S IT FOR TODAY

SPONSOR US

Get your product in front of over 5k+ readers

Our newsletter is read by top VCs, founders, engineers, managers, and tech professionals in the US. Book a slot →

FEEDBACK

Reply to this email, I would love to hear from you.

Cheers,

Chief AI Officer (CAIO)

Reply

or to participate.