AI fails differently than normal software. To make sense of it, Notion, Ramp, and Stripe use Braintrust to run thousands of evals a day and ship updates within 24 hours.

Braintrust sits between your app and your models to bring evals and observability together in one workflow. Teams use it to:

1️⃣ Define what “good” is and measure against it

2️⃣ See what happens in production

3️⃣ Connect evals and observability into a continuous improvement loop

Start shipping quality AI at scale

Anthropic and OpenAI both announced separate enterprise AI ventures backed by major financial firms, with Anthropic's valued at $1.5B and OpenAI's targeting a $10B valuation.

OpenAI was seeded by an offshoot of Y Combinator called YC Research in 2016, when Altman was running YC. Y Combinator owns about 0.6% of OpenAI. At OpenAI's current valuation, that stake is worth over $5 billion.

Orbit is a briefing and insights system in Claude and Claude Code that can produce personalized briefings with actionable insights drawn from connected work tools. Anthropic's Code with Claude developer conference will be held in San Francisco on May 6, London on May 19, and Tokyo on June 10. It is uncertain whether Orbit will be formally unveiled on stage or quietly rolled out.

OpenAI detailed a redesigned WebRTC architecture using a split relay and transceiver model to maintain low-latency, real-time voice interactions at global scale.

AI is rapidly approaching end-to-end automation of its own R&D, with major gains in coding, experiment execution, and long-horizon task autonomy. Benchmarks show models now handle complex engineering and scientific workflows, manage other agents, and increasingly outperform humans on key subproblems. If trends hold, there's a ~60% chance of self-improving AI systems by 2028, leading to recursive progress, massive productivity gains, and a capital-heavy, human-light “machine economy.”

GPT-5.5 launched with a 2x price increase over GPT-5.4. The price increase is mitigated by the model generating fewer completion tokens for longer prompts. The actual cost increase is between 49% to 92%.

73% of enterprises say data connectivity is their #1 barrier to scaling AI. Join CData and Microsoft on May 13th for the data architecture blueprint for AI agents. You'll get a framework to move from pilots to agents that act as colleagues on complete business context. Register here

Deepsec is an agent-driven security tool that scans large codebases locally or in parallel cloud sandboxes to uncover complex vulnerabilities.

The Gemini API now supports event-driven Webhooks. The push-based notification system eliminates the need for inefficient polling. The feature is available now for all developers using the Gemini API.

Tuna-2 outperforms both Tuna-R and Tuna across a diverse suite of multimodal benchmarks by using pixel embeddings. Meta plans to only release a foundation checkpoint rather than the full production-trained model weights. The release will have a small number of layers removed from both the LLM backbone and the diffusion head, but the remaining layers and all other components are fully preserved. Examples of images generated by the model are available in the repository.

ChatGPT's viral "smile" retention curve obscured a monetization gap because it tracked gross rather than net retention, with even the most engaged consumers capped at $20/month while Anthropic's $44B B2B revenue grows on per-user spend expansion. Consumer AI fails to capture value the way coding agents and legal AI do because users don't view answers or fun images as worth paying for and resist coughing up subscription dollars for savings they already pocket.

Bustamante dissects Codex CLI, Claude Code, and GitHub Copilot CLI to show that frontier labs post-train models against specific harnesses, baking tool names, schemas, citation tags, memory rituals, and system prompt structures into the weights. Terminal-Bench 2.0 data backs the thesis: Claude Opus 4.6 scored 79.8% with ForgeCode versus 75.3% with Capy, and Cursor jumped from "Top 30 to Top 5" by changing only the harness, while OpenAI models default to patch-based file edits and Anthropic models to string replacement, with mismatches costing reasoning tokens.

Raw latency doesn't tell you if the answer was right. Learn the metric that actually matters in production.

Read the guide.

DigitalOcean AI-Native Cloud is a purpose-built platform for the inference and agentic era that integrates five layers from silicon to agents into a single open stack.

An end-to-end pipeline jointly optimized image tokenization and generation, enabling direct feedback from generation quality.

AI's subtle distortion of written language has the potential to affect cultural institutions.

The Trump administration is discussing a potential executive order to create an AI working group that would bring together tech executives and government officials to examine potential oversight procedures.

TLDR is looking for an engineer/researcher at a major AI lab or startup to help write for 1M+ subscribers. Our curators have been invited to Google I/O and OpenAI DevDay, scouted for Tier 1 VCs, and get early access to unreleased TLDR products. Learn more.