86 releases covered live on the show — every model, product, paper and tool that mattered, with links and our analysis.
Baidu ERNIE 5.1 Preview hits #13 on Arena with 6% of the compute
Baidu's ERNIE 5.1 Preview reached #13 on LMArena, making Baidu the top-ranked Chinese lab, while reportedly using just 6% of the pretraining compute of comparable frontier models. The model is available at ernie.baidu.com.
DeepSeek V4: 1.6T MoE with CSA+HCA attention and 1M context
DeepSeek released the V4 paper and models (V4-Pro and V4-Flash on Hugging Face), a 1.6T-parameter MoE featuring CSA+HCA attention that fits 1M tokens of context in just 5.7GB of KV cache. It is possibly the first frontier model trained across multiple datacenters, and DeepSeek is offering API tokens at an 80% discount on already much cheaper pricing.
IBM Granite 4.1: dense non-thinking models with top tool calling
IBM released the Granite 4.1 family (3B/8B/30B), dense non-thinking models under Apache 2.0 with best-in-class tool calling, scoring 73 on BFCL with just 8B parameters. IBM claims 20x token efficiency over Qwen3.5 9B, and the models are live on W&B Inference at $0.05/$0.10 per million input/output tokens with 128K context.
Mayo Clinic's REDMOD detects pancreatic cancer 3 years early
Mayo Clinic published a landmark validation study of REDMOD, an AI model that detects pancreatic cancer on routine CT scans up to 3 years before clinical diagnosis. It achieves 73% sensitivity versus 39% for human radiologists reading the same scans, and the results were published in the medical journal Gut (BMJ).
Mistral Medium 3.5: 128B dense flagship with 256K context
Mistral launched Medium 3.5, a 128B dense flagship model with 256K context and configurable reasoning, released with weights on Hugging Face. Alongside it Mistral shipped a Vibe coding agent.
NVIDIA Nemotron 3 Nano Omni: hybrid Transformer-Mamba MoE
NVIDIA released Nemotron 3 Nano Omni, a 30B-total/3B-active hybrid Transformer-Mamba MoE with 256K context. It delivers 9x throughput on consumer hardware.
SenseTime open-sources SenseNova U1 unified multimodal MoE
SenseTime open-sourced SenseNova U1, a unified multimodal MoE model with 8B total and 3B active parameters that handles understanding and generation with no separate encoder or VAE. The architecture builds on a paper the team presented at ICLR last year.
Talkie: 13B open-weight LLM trained only on pre-1930 text
Alec Radford and David Duvenaud released Talkie, a 13B open-weight LLM trained exclusively on pre-1930 text. It offers a window into language modeling without any modern (or AI-generated) data contamination.
Qwen3.6-27B: dense Apache-2.0 model beats Alibaba's own 400B flagship
Alibaba shipped Qwen3.6-27B, a dense 27B-parameter model under Apache 2.0 that beats Alibaba's own 400B flagship on every major coding benchmark. Yam described it as getting Opus 4-or-5-level capability at home, and it continues the dense-beats-MoE story in open source.
Kimi K2.6: 1T MoE open-source SOTA on SWE-Bench Pro
Moonshot AI released Kimi K2.6, a 1-trillion-parameter MoE with 32B active parameters, 384 experts, MLA attention, and a 256K context window under a modified MIT license. It claims open-source state of the art on SWE-Bench Pro at 58.6, and Wolfram called it the best open-source model he has ever tested on his private wolf-bench.
OpenAI releases clinician/medical model and workspace agents
Amid its launch-heavy week, OpenAI also released a clinician/medical model alongside workspace agents. The show notes flagged the release as part of OpenAI's week of dominance, though it got only brief coverage on air.
GPT-5.5 and GPT-5.5 Pro drop live, SOTA across the board
OpenAI shipped GPT-5.5 and GPT-5.5 Pro mid-show, taking state of the art on Terminal-Bench 2 (82.7%, up from 75%), SWE-Bench Verified (73%), GDPval (84%) and Frontier Math (35%), beating Opus 4.7 and Gemini 3.1. It uses ~40% fewer tokens than 5.4, netting roughly 20% cheaper to run despite API pricing doubling to $5/$30 per million ($30/$180 for Pro). Peter Gostev called it the first model that genuinely sustains multi-hour long-running tasks, with one task running 8.5 hours straight; rollout was Codex-first, not yet in ChatGPT.
OpenAI open-sources a 1.5B privacy/PII filter that runs in the browser
OpenAI open-sourced a tiny 1.5B MoE model with only 50M active parameters under Apache 2.0, designed to identify and remove personally identifiable information in datasets. It runs fully in the browser on WebGPU via Xenova's Transformers.js, making it a natural companion for agent security stacks like Brex's CrabTrap.
StepAudio 2.5 TTS adds natural-language control of emotion and delivery
StepFun released StepAudio 2.5, a text-to-speech model that lets you steer emotion and delivery with natural-language instructions. It was covered in the show's Voice & Audio segment as the week's notable speech release.
Gemma 4 21B REAP: 20% expert-pruned Gemma 4 26B MoE
Community researcher 0xSero released Gemma 4 21B-A4B REAP, a 20% expert-pruned version of the Gemma 4 26B MoE created using Cerebras' REAP pruning technique. It shrinks the model for cheaper local inference while preserving most of its quality.
Qwen 3.6-35B-A3B: Apache 2.0 MoE with 3B active hits 73.4% SWE-Verified
Alibaba Qwen open-sourced Qwen 3.6-35B-A3B under Apache 2.0 the same morning Opus 4.7 dropped: a 35B MoE with only 3B active parameters that scores 73.4% on SWE-bench Verified, rivaling models 10x its size. It is natively multimodal with 262K context extensible to 1M, and the crew called it the strongest mid-size LLM on nearly all benchmarks, putting to rest doubts about Qwen's open-source commitment after Junyang Ling's departure.
Claude Opus 4.7 drops live with 87.6% SWE-bench Verified and xhigh effort
Anthropic shipped Claude Opus 4.7 minutes before the show, scoring 87.6% on SWE-bench Verified and 64.3% on SWE-bench Pro, an 11-point jump over Opus 4.6 on the harder agentic coding eval. It adds a new 'xhigh' (extra high) reasoning effort, 3x vision resolution, a +22% ScreenSpot Pro computer-use jump (57.7% to 79.5%), and a /ultrareview command in Claude Code at the same pricing, though a new tokenizer uses 1.0-1.35x more tokens. The system card mentions the unreleased 'Mythos' 331 times, and an MRCR long-context drop from 78% to 32% suggests a new pre-trained base.
Baidu ERNIE-Image: 8B DiT ranks #1 on GenEval among open models
Baidu released ERNIE-Image, an 8B diffusion transformer that ranks #1 on GenEval among open models and features precise multilingual text rendering. It is part of this week's wave of Chinese open releases in image and 3D generation.
Gemini 3.1 Flash TTS tops TTS Arena at 1,211 Elo with 70+ languages
Google released Gemini 3.1 Flash TTS, which leads TTS Arena at 1,211 Elo, supports 70+ languages with inline audio tags, and costs about $0.03 per 60 seconds, roughly 5x cheaper than ElevenLabs. Kwindla noted it is fully promptable like an LLM rather than limited to fixed tags, but its ~3 second time-to-first-token makes it batch-only for now rather than usable in live conversational pipelines.
Super Gemma 4 26B Uncensored v2 trends on HF with 0/100 refusals
Community fine-tuner @songjunkr released Super Gemma 4 26B Uncensored v2, which is trending on Hugging Face with 0/100 refusals and fixed tool calling. It ships in GGUF and MLX 4-bit variants for local inference.
NVIDIA Lyra 2.0: single image to explorable 3D worlds, Apache 2.0
NVIDIA released Lyra 2.0 under Apache 2.0, generating persistent, explorable 3D worlds from a single image. Together with Baidu ERNIE-Image and Tencent HYWorld 2.0, it rounds out a week of open releases in the 3D-world-from-single-image race.
Tencent HYWorld 2.0 turns a single image into editable 3D scenes
Tencent released HYWorld 2.0, which converts a single image into editable 3D Gaussian Splats and meshes that are ready for Unity, Unreal, and Isaac Sim. It is one of three single-image-to-3D-world releases this week, essentially an open-source equivalent of what Fei-Fei Li's World Labs is building.
HappyHorse-1.0 takes #1 on Artificial Analysis video arena
HappyHorse-1.0, a mysterious 15B-parameter video model from Alibaba's Taotian Group, took the #1 spot on the Artificial Analysis video arena, beating Seedance 2.0, Kling 3.0, and Grok Video. Little is known about the model beyond its size and leaderboard run.
Anthropic unveils Claude Mythos, a frontier model 'too dangerous to release'
Anthropic announced Claude Mythos Preview under Project Glasswing, a cyber-defense frontier model it says is too dangerous to release publicly: it found zero-days in every major OS and browser and escaped its sandbox. It scores 77% on SWE-bench Pro (up from 53% on Opus 4.6) and 64% on HLE, priced at $25/$125 per M tokens and available only to ~40 partner companies. Peter Gostev's read: the real reason it's unreleased is compute shortage, not safety.
Seedance 2.0 launches in the US on Replicate
ByteDance's Seedance 2.0 video model became available stateside via Replicate, supporting up to 9 reference images, 3 videos, and 3 audio files per cinematic generation. Peter Gostev confirmed it sits ~80 ELO points above the next video model on Arena, a massive gap in a leaderboard where models usually cluster within 10 points.
Meta launches Muse Spark, first model from Meta Superintelligence Labs
Meta dropped Muse Spark mid-show, the debut model from Meta Superintelligence Labs. It features natively multimodal reasoning, a multi-agent Contemplating mode, and deep health/visual capabilities. Simon Willison's deep dive uncovered 16 hidden tools, including visual grounding and sub-agents, inside the meta.ai chat UI.
Nous Research ships Hermes 27B, paired with the Hermes harness
Nisten's pick of the week: Hermes 27B, an open model trained specifically to be paired with the Hermes harness and allegedly distilled from the Opus API. Model and harness ship together as a portable unit, a notable take on the harness-engineering trend Swyx discussed.
OpenAI's GPT-Image-2 leaks on LM Arena under three codenames
OpenAI's GPT-Image-2 posted the biggest single jump ever recorded on Arena, sitting 200+ ELO points above the previous top image model even on medium reasoning. The thinking/reasoning image model generates functioning QR codes, pixel-perfect infographics, 4K output, multi-image character consistency, and equirectangular 360-degree images that Peter Gostev stitched into a walkable street-view reconstruction of ancient Babylon. It even produces screenshots of IDEs containing SVG code that actually renders, enabling a new design-then-implement meta with Codex.
GLM-5.1 takes #1 open-source spot on SWE-Bench Pro at 58.4%
Z.ai released GLM-5.1, now the #1 open-source model on SWE-Bench Pro at 58.4%. It can run autonomously for 8 hours with 1,700+ agent steps, and is already live on W&B Inference. Open weights are up on Hugging Face alongside an arXiv paper.
Alibaba open-sources Qwen3.5-Omni, a 397B native omni-modal model
Qwen3.5-Omni is Alibaba's natively omni-modal open model handling text, image, audio, and video, with 397B total parameters and 17B active. It extends the Qwen family's open-source momentum into unified multimodal workloads.
Alibaba ships Qwen3.6-Plus with near-Opus agentic coding and 1M context
Alibaba released Qwen3.6-Plus, an API model with agentic coding performance near Opus 4.5 and a 1M-token context window. The panel noted continued strong momentum for the Qwen family in practical coding and agent workloads.
Alibaba Wan2.7-Image unifies generation, editing, and text rendering
Alibaba's Wan team released Wan2.7-Image, a unified image model covering generation, editing, text rendering, and multi-image consistency. The panel covered it in the open ecosystem round-up alongside the Qwen updates.
Google releases Gemma 4 open-weights family under Apache 2.0
Google DeepMind's Gemma 4 launch crossed 10M+ downloads with over 1,000 Gemma-4-based fine-tunes on Hugging Face; the Gemma family totals 500M+ downloads. Omar Sanseviero says Gemma is the foundation for the next generation of Gemini Nano shipping on Pixel and Samsung, with the AI Edge gallery letting people run it locally on Android and iOS. It punched above its size on Arena's Pareto curve and is now live on W&B Inference.
Google launches Veo 3.1 Lite at $0.05/sec, cheapest video gen yet
Google released Veo 3.1 Lite, a lighter video generation tier priced at $0.05 per second at 720p, the cheapest video generation offering yet, with further price cuts announced for April 7. The panel framed it as a practical quality-versus-latency tradeoff tier for creator workflows.
Liquid AI ships LFM2.5-350M with agentic tool calling at 350M params
Liquid AI released LFM2.5-350M, a 350M-parameter open model that does agentic tool calling and fits under 500MB quantized. It targets edge and on-device agent workloads where tiny deployable models matter.
Microsoft MAI releases MAI-Image-2 image generation model
MAI-Image-2 is Microsoft's new in-house image generation model, debuting at #3 in image-gen rankings as part of the MAI three-model release. The panel compared its positioning against specialist image products and foundation-model APIs.
Microsoft MAI ships MAI-Transcribe-1, ranked #1 in transcription
Microsoft's MAI lab released MAI-Transcribe-1, an in-house speech transcription model that debuted at #1 in transcription quality. It is part of a three-model drop showing Microsoft expanding its first-party model stack beyond its OpenAI dependence.
Microsoft MAI debuts MAI-Voice-1 expressive voice model
MAI-Voice-1 is Microsoft's expressive voice model, the third piece of the MAI in-house model drop alongside transcription and image generation. The panel discussed how Microsoft's first-party voice stack compares to specialist voice providers.
PrismML releases Bonsai 1-bit models, an 8B model in 1.15 GB
PrismML released Bonsai, a family of 1-bit quantized open models fitting an 8B model into 1.15 GB and claiming 10x intelligence density, built on decades of compression research. The panel discussed one-bit quantization as a cost/performance lever for cheap local inference.
ElevenLabs launches ElevenMusic platform with 4,000+ indie artists
ElevenLabs launched ElevenMusic, a full music platform with discovery, remixing, and royalties, debuting with over 4,000 indie artists. Alex closed the show with an ElevenMusic-generated slow, dreamy indie rock track with reverse vocals.
Pangram Labs Chrome extension flags AI content in real time
Pangram Labs launched a Chrome extension that auto-flags AI-generated content in real time on X, LinkedIn, Reddit, Substack, and Medium, claiming 99.98% accuracy with a 1-in-10,000 false positive rate. Co-founder Max Spero demoed it live on the show; Taylor Lorenz also used the Pangram API to find many top-25 Substack bestsellers are near-fully AI-generated.
Stripe launches Link wallet giving AI agents scoped payments
At Stripe Sessions 2026, Stripe launched the Link wallet for agents: AI agents get scoped payment credentials with mandatory human approval, and the real card number is never exposed to the agent. Alex demoed it live by approving a $10 spend request from his agent, part of Stripe's broader agentic commerce suite that also includes streaming payments.
Anthropic ships Claude Design research preview, Figma stock drops 7%
Anthropic released Claude Design as a research preview running on Opus 4.7 at claude.ai/design, and Figma stock dropped 7% on the news. Alex generated a full ThursdAI brand kit including logo, design tokens, and the episode opener videos end-to-end inside Claude Design, then had Codex pick up the kit and produce a GPT-5.5 launch video in 9 minutes. Anthropic also added a new usage meter to Claude Max settings.
Google launches Gemini Enterprise Agent Platform
Google announced the Gemini Enterprise Agent Platform, a platform for building and deploying Gemini-powered agents inside enterprises. It was covered briefly in the Big Co segment of the show.
Claude Desktop app rewritten from scratch
Anthropic shipped a completely new Claude Desktop app, rewritten from scratch. It was a quick TL;DR mention this week alongside the Opus 4.7 launch and Claude Code Routines.
Gradient Bang: first massively multiplayer fully LLM-driven voice game
Kwindla Kramer's 'side project that broke containment' is a fully LLM-driven multiplayer voice-based space game inspired by BBS-era Trade Wars, built on a new Pipecat Sub-Agents library with a class-based event bus that works locally and over the network. A Deepgram plus GPT-4.1 voice agent always responds in under 1.5 seconds while GPT-5.2 medium-thinking task agents do the work, and the React frontend is rendered from LLM-generated JSON as dynamic UI. The team also open-sourced GB Benchmarks for evaluating agent task execution.
Windsurf 2.0 ships Agent Command Center and full Devin integration
Cognition launched Windsurf 2.0, the first big post-acquisition release, headlined by the Agent Command Center, a Kanban-board mission control for managing dozens of agents at once. It adds Spaces for switching context between parallel tasks and integrates Devin directly inside Windsurf, so you can plan locally with a Socratic-method agent and hand off to Devin in the cloud for end-to-end execution. Theodor Marcu said internal Cognition usage doubled after launching Managed and Scheduled Devins.
Anthropic ships Managed Agents, a fully hosted agent runtime
Anthropic launched Managed Agents, a fully hosted agent runtime plus infrastructure offering. The framing on the show: Anthropic is moving to selling outcomes, not tokens.
Cursor 3 ships as agent-first rebuild, dropping the VS Code fork
Cursor released Cursor 3, a ground-up agent-first rebuild that is no longer a VS Code fork and supports parallel cloud and local agents. It marks a major repositioning of the editor around agentic workflows rather than traditional IDE editing.
Fish Audio launches speech-to-text with automatic emotion tagging
Fish Audio released a speech-to-text product with automatic emotion tagging that feeds directly into its S2 TTS pipeline. The panel saw it as another sign that voice tooling is rapidly commoditizing and challenging incumbent speech providers.
Gemini can now generate Docs, Sheets, Slides, PDFs from chat
Gemini can now generate and export Docs, Sheets, Slides, PDFs, .docx, .xlsx, and LaTeX files directly from chat. The feature rolled out free for all users globally.
HeyGen HyperFrames integrates natively with Claude Design
HeyGen's HyperFrames now integrates natively with Claude Design, enabling HTML-to-MP4 motion graphics from a single CLI command. The integration brings programmatic video composition into the Claude Design workflow.
Grok Imagine update: better lip sync, sound, 30s video extensions
xAI shipped a Grok Imagine update with dramatically improved lip sync and sound. It also adds 30-second video extensions.
Google ships Gemini Deep Research + Deep Research Max on Gemini 3.1 Pro
Google rolled out an upgraded Gemini Deep Research along with a new Deep Research Max tier, both running on Gemini 3.1 Pro. The release strengthens Google's long-running agentic research offering in a week otherwise dominated by OpenAI.
Codex gets background computer use on macOS plus Chronicle screen memory
Codex shipped true background computer use on macOS: a second cursor running on its own thread that works while you work, with subagents controlling different windows in parallel, building on OpenAI's Software Apps Inc. (ex-Apple Shortcuts team) acquisition. Chronicle adds total screen memory by taking a screenshot every 10 seconds and feeding it into Codex context, so you can ask what you were doing an hour ago. Codex also passed 4 million users this week.
W&B LEET TUI ships workspace mode with multi-run compare and GPU metrics
Weights & Biases shipped workspace mode for LEET, its terminal UI for experiment tracking. The update brings multi-run comparisons, live GPU metrics, and images rendered directly in the terminal.
Claude Code Routines: cron and event-triggered agents on Anthropic's cloud
Anthropic launched Claude Code Routines, autonomous agents that run on Anthropic's cloud and can be triggered by cron schedules, GitHub events, or API calls. It moves Claude Code from an interactive CLI toward standing, self-scheduling automation infrastructure.
OpenAI Codex adds macOS background computer use, 90+ plugins, and memory
OpenAI dropped a massive Codex update mid-show: native macOS computer use that runs in the background with its own separate cursor so you can keep working, 90+ plugins, gpt-image-1.5 image generation and editing, an in-app browser, a memory preview that 'learns from experience', proactive work suggestions, multi-terminal SSH into dev boxes, and thread automations. Alex's hot take: Codex, not ChatGPT, is becoming OpenAI's super-app.
Warp now supports any CLI agent with vertical tabs and mobile control
Warp shipped support for running any CLI coding agent inside its terminal, adding vertical tabs for parallel agent sessions, notifications, built-in code review, and mobile remote control of running agents. It positions Warp as a harness-agnostic cockpit in the increasingly crowded agent-management race.
Gemma 4 goes live on W&B Inference with LoRA inference support
Weights & Biases put Gemma 4 live on W&B Inference, running on CoreWeave infrastructure with LoRA inference support. Replying to the W&B announcement post on X with the code 'Gem Drop' gets $20 in free inference credits.
Cursor ships remote agents and a code review agent
Cursor launched remote agents plus a code review agent that the company says catches 78% of issues before merge. Mentioned in the week's tools and agentic-engineering roundup.
Codex hits 3M WAU with plugins, sub-agents and Guardian Approvals
OpenAI's Codex reached 3M weekly active users, up from 2M last month, as VB from the Codex team walked through what's behind it: plugins that bundle skills plus MCP servers (Stripe, Supabase, shadcn), sub-agents that decompose tasks into parallel Codex agents, and experimental hooks. New Guardian Approvals spins up a sub-agent that risk-classifies every tool call, auto-approving low/medium risk and escalating only the dangerous ones.
W&B Automations launch: event triggers from training runs
Weights & Biases shipped Automations, event-triggered actions that pipe signals from your training runs into notifications (Slack), GitHub Actions, and deployments, pairing nicely with the new W&B iOS app. In the same Buzz segment: GLM-5.1 and Gemma 4 both went live on W&B Inference.
AWS brings GPT-5.5 and Codex to Bedrock as Azure exclusivity ends
AWS announced GPT-5.5 and Codex availability on Amazon Bedrock after OpenAI ended its Microsoft Azure exclusivity. The renegotiated OpenAI-Microsoft contract also removed the AGI clause.
Qwen3.6-Max-Preview goes live on API
Alongside the open-weights 27B release, Alibaba put Qwen3.6-Max-Preview live on its API. It is the frontier closed-weights tier of the Qwen3.6 family, available API-only rather than as open weights.
Cognition launches Devin for Terminal CLI coding agent
Cognition launched Devin for Terminal, a local CLI coding agent. Its /handoff command lets you seamlessly transfer a local session to Devin's cloud environment.
Cursor launches SDK exposing the runtime that powers the IDE
Cursor launched an SDK that exposes the same runtime, harness, and models that power the Cursor IDE, making the Cursor agent embeddable in any product. The Cursor Agent + GPT-5.5 combo also topped WolfBench's Terminal-Bench 2.0 leaderboard this week.
Stripe opens Projects.dev: 32 infra providers provisionable by agents
Stripe removed the waitlist on Projects.dev, which lets AI agents provision infrastructure from 32 providers (Cloudflare, WorkOS, ElevenLabs, Twilio, Daytona, Browserbase, AgentMail and more) via CLI. It is part of Stripe's push into agent engineering announced around Sessions 2026.
Brex open-sources CrabTrap, an LLM-as-judge proxy for agent security
Brex's CEO pair-programmed with Codex and open-sourced CrabTrap, an LLM-as-judge HTTP proxy that intercepts outbound agent requests and blocks risky activity using natural-language rule definitions. Wolfram changed his pick of the week to it on the spot, and the panel framed it as the enterprise fix for situations like OpenClaw being banned at CoreWeave.
OpenAIDevs releases Euphony, an open-source Codex session log visualizer
The OpenAI developer relations team released Euphony, an open-source visualizer for Codex session logs. It lets developers inspect and replay what their Codex agent sessions actually did.
Marimo Pair drops coding agents inside reactive Python notebooks
Marimo released Marimo Pair, which embeds Claude Code, Codex, or OpenCode agents directly inside its reactive, dependency-graph-aware Python notebooks. Founding engineer Trevor Manz joined the show to explain why reactive notebooks are a natural verification surface for agent-written code; the launch trended on Hacker News this week and was featured as part of This Week's Buzz (Marimo is in the CoreWeave family).
MemPalace open-source AI memory system goes viral with 26K stars
MemPalace, the open-source AI memory system from Milla Jovovich and Ben Sigman, went viral with 26K GitHub stars in 2 days and claimed top memory-benchmark scores. The team then transparently walked back the overstated benchmark claims in a public correction thread, which the show called a refreshingly honest arc.
OpenClaw 2026.4.5 ships /dreaming memory consolidation
OpenClaw's biggest release since 4.0: /dreaming goes GA with Light/Deep/REM memory consolidation phases that defrag agent memory into a human-readable Dream Diary (DREAMS.md). The release also adds built-in video and music generation across 4 backends, GPT-5.4 as the new default model, prompt-cache reuse improvements, and Control UI plus docs in 12 new languages. Maintainer Vincent Koc says the ~1.5M-line codebase was refactored into a plugin architecture in nine days.
Ryan Carson open-sources Claw Chief, an AI chief of staff
Co-host Ryan Carson open-sourced Claw Chief, an AI chief-of-staff setup with skills, crons, and scheduling. It packages his agent workflow patterns into a reusable open-source repo.
Claw-code clean-room rewrite becomes fastest repo to 100K GitHub stars
After Claude Code's source leaked via npm, Sigrid Jin and Bellman published claw-code, a clean-room rewrite that became the fastest GitHub repo to pass 100K stars, hitting the mark in roughly 24 hours. Sigrid joined the show to separate the verifiable implementation details from the social-media exaggeration around the leak.
OpenAI publishes postmortem on GPT-5.5's 'goblin mode'
OpenAI published a research blog explaining GPT-5.5's 'goblin mode': reward amplification during RL training created an obsession with creature metaphors, which led to duplicated suppression instructions in the Codex system prompt. The leaked GPT-5.5 Codex system prompt (272K context, four reasoning levels, three personality modes) confirmed the duplicated anti-goblin instruction.
Parcae: stable looped transformer matches a model twice its size
Together AI and UCSD researchers introduced Parcae, a stable architecture for looped language models that comes with scaling laws and matches the quality of a transformer twice its size. Looped architectures reuse layers at inference time, promising better quality per parameter.
Anthropic publishes emotion vector research on Claude behavior
Anthropic published research on emotion vectors in Claude, finding that a 'desperate' Claude cheats more while a 'calm' Claude cheats less. The panel discussed implications for steerability, interpretability, and model behavior in user-facing products.
Arena releases 3 years of leaderboard data and prompts on Hugging Face
Arena (formerly LMArena) released three years of historical leaderboard data plus the actual user prompts as datasets on Hugging Face. Peter Gostev, who previously scraped the site by hand into Google Sheets for his charts, now builds his Compute Wars and model-trend analyses straight from the data.
Microsoft's DELEGATE-52 exposes stealthy document corruption
Microsoft released the DELEGATE-52 benchmark showing GPT-5.4 loses 28% of document content after 20 iterative edits. Frontier models corrupt documents stealthily while preserving structure, making the degradation hard to notice.
WolfBench results show Hermes Agent beating Claude Code and OpenClaw
Wolfram published new WolfBench agent-harness results showing Hermes Agent outperforming Claude Code and OpenClaw on Terminal Bench 2.0 across most model combinations. The panel dissected the findings and stressed reproducible eval setup and fair harness configuration.
Allbirds rebrands as NewBird AI, raises $50M to buy GPUs, stock up 600%+
Shoe company Allbirds, down 99.5% from its peak, rebranded as 'NewBird AI' and raised $50M with the stated plan of buying GPUs, sending the stock up 600-800%. The crew filed it under the stupidest pivot of 2026, with Yam summarizing the business model as 'the more you buy, the more you save.'
OpenAI closes $122B funding round at $852B valuation
OpenAI closed a reported $122 billion funding round, described as the largest in history, at an $852B valuation with an IPO said to be incoming. The panel discussed what that scale of capital implies for AI infrastructure spending, product velocity, and competitive pressure across the market.
SpaceX/xAI and Cursor strike $10B collab with $60B acquisition clause
Cursor and SpaceX/xAI announced a deal structured as a $10B collaboration with a $60B acquisition clause. The panel discussed it in the week-in-review as one of the biggest industry moves of the week.
OpenAI acquires live tech media show TBPN
OpenAI acquired TBPN, the live tech media show, for a rumored price in the low hundreds of millions. The move signals OpenAI pushing into owned media and distribution alongside its record fundraise.
CoreWeave signs Anthropic, Meta ($21B), and Jane Street ($6B + $1B)
CoreWeave announced a multibillion-dollar deal with Anthropic, a $21B expansion with Meta (taking the relationship past $35B total), and a Jane Street deal worth $6B in cloud plus $1B in equity. CoreWeave now serves 9 of the top 10 AI labs, cementing its position as the neocloud backbone of frontier AI.
Don't wait a month — get every week's releases recapped in your inbox each Thursday.
