OpenAI Introduces GPT-5.3-Codex-Spark Model

#1 📝 OpenAI News

Introducing GPT-5.3-Codex-Spark - Announces the GPT-5.3-Codex-Spark product release, highlighting new Codex-powered capabilities for developers and product teams. The post introduces the model and its intended use cases and availability.

Also covered by: @Simon Willison

#2 𝕏

Demis Hassabis rolled out Gemini 3’s new “Deep Think” mode for Google AI Ultra subscribers in the Gemini App, enabling more advanced reasoning and complex problem-solving capabilities.

Also covered by: @Josh Woodward, @Demis Hassabis, @Google AI, @Sundar Pichai, @Sundar Pichai

#3 𝕏

Sam Altman launched GPT-5.3-Codex-Spark as a research preview for Pro today, delivering over 1,000 tokens per second with initial limitations that will be rapidly improved.

Also covered by: @Simon Willison

#4 𝕏

Josh Woodward used Gemini 3 Deep Think to turn a laptop-stand sketch into an interactive prototyping tool, generate an STL file, and 3D-print the final design via @fleet_ai.

#5 𝕏

Philipp Schmid updated Gemini 3 Deep Think in GeminiApp (now live for Ultra subscribers with API access rolling out), achieving 48.4% on Humanity’s Last Exam (no tools), 84.6% on ARC-AGI-2, a 3455 Codeforces Elo, and IMO 2025 gold-medal level.

Also covered by: @Josh Woodward, @Demis Hassabis, @Google AI, @Sundar Pichai, @Sundar Pichai

#6 📝 Simon Willison

Anthropic raises $30 billion series G - Anthropic announced a $30 billion Series G, claiming Claude Code's run-rate revenue has grown to over $2.5 billion and that weekly active users doubled since January 1. The announcement highlights rapid commercial growth for their coding product.

#7 𝕏

Cursor previews its early research on very-long-running coding agents that persist project state across sessions, self-debug, and automate extended multi-step coding workflows.

#8 ▶️

24/7 Claude Code AI Agent 12-Day Review: The Results Will Surprise You

All About AI

#9 📝 PromptLayer Blog

Understanding Intermittent Failures in LLMs - This article explains why deployed LLM applications sometimes begin returning nonsense, timeouts, or refusals despite passing tests, drawing on PromptLayer's production observations. It frames intermittent LLM failures as a common, hard-to-diagnose problem teams regularly encounter in production.

#10 📝 PromptLayer Blog

Opus 4.6 — PromptLayer Team Review - PromptLayer's team reviewed Claude Opus 4.6 after extensive testing across coding workflows, long-document analysis, and agentic pipelines. The article shares the team's verdict and insights about how the release performs in real-world engineering scenarios.

#11 📝 Eleanor Berger & Isaac Plath

Automating Presentation Slides with Agent Skills - Demonstrates creating presentation slides agentically using Slidev, Nano Banana, and Agent Skills. Presents an automated workflow for building slides with agent tools.

#12 𝕏

Andrew Ng launched A2A: The Agent2Agent Protocol, a short course built with @googlecloudtech and @IBMResearch and taught by Holt Skinner, @ivnardini, and Sandi Besen.

#13 𝕏

There's An AI For That launched Remix, an SDK that turns any React Native app into a plain-English, user-customizable experience—no forks or code editors needed.

#14 𝕏

Santiago: warpdotdev’s Oz is a cloud-based coding agent orchestration platform with a unified dashboard to spin up and manage local, cloud, scheduled, and API-triggered agents in isolated, repo-connected environments — and even fork them locally.

#15 𝕏

Boris Cherny credits Claude Code for driving their latest raise, with weekly active users doubling since January as non-coders start building with it.

#16 𝕏

DeepLearning.AI unveiled Moonshot AI’s Kimi K2.5, a vision-language model that spins up parallel workflows for coding, research, web browsing and fact-checking, then merges the outputs into a single answer.

#17 ▶️

Give Me 20 Minutes, I'll Make You AI Native

Peter Yang

#18 𝕏

Philipp Schmid breaks down why engineering teams struggle building AI agents—citing fragmented orchestration, missing CI/CD for prompts and workflows, and poor observability—and outlines targeted tooling and process fixes to speed up reliable, scalable launches.

#19 📝 Jesse Vincent

Letting agents post on my blog; finding a needle in a haystack - The author explains their prior caution about letting AI agents write blog posts except in explicitly flagged sections, and begins to discuss an experience or reasoning around allowing agents to post to their blog. The post reflects on the challenges of discerning agent-written content and finding valuable contributions.

#20 𝕏

DeepLearning.AI warns that AI beginners often fail by fixating on “Which model should I use?” before identifying a real problem—great AI starts with solving genuine needs, not chasing the latest architecture.

#21 𝕏

LlamaIndex 🦙 unveiled Long Horizon Document Agents—autonomous agents that tackle complex document workflows end-to-end over weeks.

#22 𝕏

Andrej Karpathy congratulates @simile_ai on its launch and praises their novel approach of treating a pretrained LLM as a simulation engine for entire populations of internet personas instead of a single crafted character.

#23 in

🗞️ Carl Vellotti hosted his first live Cursor workshop for 100 Seattle PMs, uncovering Claude as the top‐ranked LLM, under 10% with prior Cursor experience (25% had tried Claude Code), and a surge in companies greenlighting AI tools this year.

#24 𝕏

Santiago runs 4–6 coding agents every day (Claude Code, Copilot, Warp, Jules) but can’t keep track of who’s doing what or maintain context, realizing that his own attention, not the models, has become the bottleneck.

#25 in

Udi Menkes spotlights Levelsio’s OpenClaw agent, which bootstrapped itself by selling CloudBot skills via a landing page and has already earned $12.93.