Image via TechCrunch
Tuesday, February 17, 2026
Brain-inspired AI lab just landed $180M
Matt Wolfe just dropped his 15-tool AI stack for entrepreneurs (yikes, that's a lot of tabs), while a bold new $180M lab called Flapping Airplanes is rethinking AI training entirely by ditching bio-inspiration for pure data efficiency. On the capabilities front, Claude's decompilation prowess is hitting the long tail of real engineering problems, yet researchers are still arguing AGI is decades away because current LLMs lack embodied cognition. And if you want agents everywhere, Manus just brought full-featured AI agents to Telegram with literally one click. Would you actually use 15 different AI tools daily?
Image via TechCrunch
Top Stories
Matt Wolfe's AI Playbook provides entrepreneurs with a curated toolkit of 15 AI tools to integrate into business operations for improved efficiency, creativity, and growth.
TechCrunch
Flapping Airplanes is pursuing radically different AI architectures focused on data efficiency, drawing inspiration from neuroscience but avoiding biomimicry, targeting 1000x improvements in data efficiency to unlock new capabilities in reasoning, out-of-distribution generalization, and domains currently limited by data constraints.
Chris Lewis Blog
Using Claude to decompile legacy code hits a long tail of diminishing returns, but specialized tooling, similarity-based task scheduling, domain-specific skills, and robust agent orchestration can push completion rates to 75% while managing token costs and preventing hallucination-induced regressions.
dlants.me
Despite CEO proclamations of imminent AGI, current LLMs fundamentally lack embodied cognitive primitives and rely on compute-intensive external scaffolding rather than genuine reasoning, with leading AI researchers agreeing that scaling alone cannot bridge the gap to general intelligence.
Manus
Manus launches chat-based AI agents in Telegram with zero-setup deployment, enabling full task execution and multi-modal inputs directly in messaging apps. This shifts personal AI from isolated applications to ambient accessibility within users' existing communication tools.
Keep Reading
Industry Voices
Yann LeCun
NYU Center for Data Science
Delivers unfiltered takes on AI hype cycles and architectural debates that other Meta researchers won't say publicly.
Alexander Embiricos
Product Lead for Codex at OpenAI
Ships insights from the trenches of making AI actually write code that doesn't break production.
Joanne Jang
Founding Leader of Model Behavior Team at OpenAI
Reveals how OpenAI actually decides what ChatGPT refuses to do and why those guardrails keep changing.
Aidan Smith
Co-founder at Flapping Airplanes
Shares builder-focused perspectives on applying AI to real aerospace engineering problems beyond chatbots.
Enjoyed this issue?
Get daily AI intel delivered to your inbox. No fluff, just the stories that matter.