Thursday, May 28, 2026

Devin Lands A Billion-Dollar Boost

Devin Lands A Billion-Dollar Boost

Today’s Overview

Good morning, AI’s feedback loop is getting a serious startup push, Cognition just put a huge new valuation behind Devin, and OpenAI’s nonprofit arm is putting real money toward the economic fallout of automation. There’s also a sharp Stanford study on hiring algorithms that is worth reading closely. Let’s dive in.

Top Stories

OpenAI Foundation Puts $250M Toward AI Disruption

The OpenAI Foundation committed an initial $250 million to grants, partnerships, and direct work focused on workers, communities, and economies facing AI-driven disruption. The effort is aimed at measuring AI’s economic effects, supporting people through near-term labor shifts, and exploring longer-term models for shared economic security.

  • The program is organized around three workstreams: understanding the shift, supporting the transition, and building long-term economic security.
  • The Foundation says existing economic statistics may miss AI’s value if it shows up as digital goods instead of wages, margins, prices, tax revenue, or capital returns.
  • It also calls for stronger labor-market infrastructure, including BLS-like and O*NET-like systems that can track work, transitions, firm behavior, and local differences globally.

Trajectory Wants AI To Learn From Use

Trajectory launched with $15 million to build a platform for continual learning, where AI products improve from real-world user feedback rather than staying frozen after training. The startup captures corrections, retries, and edits from product data, then uses that signal to keep post-training models over time.

  • The seed round values the company at $115 million post-money and includes Conviction, Bessemer Venture Partners, Radical VC, BoxGroup, Jeff Dean, and Fei-Fei Li.
  • Trajectory’s founding team includes researchers from Google DeepMind and Apple with experience in coding agents, Vision Pro work, and robotics.
  • Its current workflow starts with open-source models that are post-trained for specific customer products rather than defaulting to off-the-shelf frontier models.

Cognition Raises $1B At $26B Valuation

Cognition, the creator of Devin, raised more than $1 billion at a $26 billion valuation. The company says enterprise usage has grown more than 10x since the start of the year, driven by demand for its AI software engineer.

  • The round was led by Lux Capital, General Catalyst, and 8VC with support from existing and new investors including Founders Fund, Elad Gil, Ribbit Capital, Atreides, and Layer Global.
  • Cognition says its run-rate revenue has reached $492 million as cloud agents move from niche tools into mainstream software development workflows.
  • The company says 89% of code committed by its own engineers is committed by Devin, with the rest coming from local agents in Windsurf.

Research & Analysis

Stanford Study Finds Bias In AI Hiring

Stanford researchers analyzed millions of job applications across 156 employers and found clear racial disparities in algorithmic hiring outcomes. The study is especially concerning because shared models can create algorithmic monocultures, where the same applicants or demographic groups are repeatedly screened out across employers.

  • The dataset covers 4,197,168 applications from 3,372,132 applicants across 1,746 positions screened by the same vendor.
  • The authors found that 25.87% of applications submitted by Black applicants went to positions with adverse impact against Black applicants under U.S. employment discrimination standards.
  • Applicants would need to submit 25 applications to bring systemic rejection below 0.1%, compared with 10 under a baseline where employers made independent decisions.

Biohub Releases Protein Biology World Model

Biohub released new Evolutionary Scale Models for mapping, predicting, and designing proteins. The stack includes ESMFold2, ESMC, and ESM Atlas, creating an open foundation for protein structure prediction, binder design, and molecular discovery.

  • ESMC was trained on approximately 2.8 billion sequences drawn from across life to learn protein representations used by the broader system.
  • ESM Atlas maps 6.8 billion sequences and 1.1 billion predicted structures, making protein space searchable at massive scale.
  • Biohub says ESMFold2 designs produced 36% to 88% hit rates for compact minibinders in lab experiments, with antibody-derived formats reaching 15% to 29%.

Trending AI Tools

  • Robinhood Agentic Trading Lets users connect AI agents to execute stock trades, manage spending, and automate purchases.

  • Google Coral Board A low-power development platform powered by Coral NPUs for on-device AI applications.

  • Buffer API One API for publishing across social platforms.

Quick Hits

Keep reading for free

Enter your email. If you're already subscribed, we'll send a sign-in code. If not, you'll subscribe in the next step.

Free access. Subscribe once, then use the same email on future issues.

Free to read. Subscription just unlocks the full issue.