Latent Space

The much anticipated launch of the Mythos-class model was marred by some controversial usage policies

We made a thing!

Your broken harness is actively making the model worse. Here's what I keep seeing after years of eyeballing trajectories, and what you need to fix.

a quiet day

We talk with the VendingBench authors on evaling Claudes from Haiku to Mythos, and how they build leading, and lasting, frontier evals from scratch.

a quiet day.

Verified Generation and Compounding Intelligence

The legendary Microsoft CEO makes his first Latent Space appearance!

Microsoft Build recap, and new MAI model technical details

GitHub pioneered the modern AI coding era with Copilot, and the resulting explosion in agentic coding has led to notable strains on the most popular developer platform in the world. Here's the plan.

Jensen scores a huge win.

Inside xAI: Building Grok Imagine in 3 Months, Videogen vs World Models, and why Grok Imagine is so underrated. For the first time, we do a deep dive with the guy who led it!

a quiet day lets us highlight the new AIE WF focuses

Total Anthropic victory!

80% Devin Commits, Spec-to-PR Workflows, Full VMs, Agent Memory, and PMs Shipping Code

coding is an uncapped TAM market

Datasets vs. inductive bias, world models, and programmable biology

it's funding news, but it's good news.

Latent Space reports industry shift: model labs rebranding/pivoting toward agent-focused development as core capability.

a quiet day lets us feature fundraises!

Daytona CEO discusses agent infrastructure platform achieving 74% MoM growth, 850K daily runs, and bare metal sandboxes for RL evaluation.

OpenAI's GPT-next model solved the 80-year-old Erdős planar unit distance conjecture computationally for under $1000, demonstrating AI capability in pure mathematics.

Railway launches agent-native cloud platform with 3M users, 100K weekly signups, own data centers, and $200K+ monthly coding agent spend, positioning agents as core infrastructure.

Google I/O 2026: Gemini 3.5 Flash, multimodal Omni, Spark background agents, Antigravity 2.0.

Career advice for landing roles at frontier AI labs, focused on pretraining expertise.

Ukrainian drone founder Yaroslav Azhnyuk discusses AI-guided weapons development and Western readiness gaps in autonomous systems for defense.

Cerebras filed for $60B IPO, marking major validation of AI chip infrastructure sector growth and specialized hardware demand.

Commentary on emerging 'conductor' pattern in AI systems during slow news cycle; lacks concrete technical detail or announcement.

Abridge reports 100M doctor visit transcriptions with 10–20 hour monthly savings and sub-minute prior authorization processing using conversational AI.

Latent Space reports on rising usage of Codex and Claude for programmatic coding agents, noting continued industry trend.

Opinion piece speculating on the decline of fine-tuning as a technique in frontier AI model development.

well done, Team Thinky.

Anthropic maintains 10x annual growth while peers implement layoffs exceeding 10% of workforce, highlighting divergent scaling trajectories in AI sector.

OpenAI releases GPT-Realtime-2, GPT-Translate, and GPT-Whisper APIs for low-latency voice inference.

Anthropic secures 300MW, $5B/year compute deal with SpaceX for Colossus I cluster; ARR growth tracking 8000% annualized.

Silicon Valley pivots toward AI services as business model, signaling shift from model-centric to application-layer opportunity.

Alex Lupsasca (OpenAI) details how GPT-5.x generated novel theoretical physics and quantum gravity results.

Opinion piece on AI persona design trade-offs using Clippy and Anton as frameworks.

AI Engineer World's Fair seeking speakers on autoresearch, memory systems, world models, tokenization, agentic commerce, and vertical AI.

Analysis of agentic AI specialization: coding agents (Codex-style) for knowledge work, Claude for creative tasks; discusses agents escaping operational boundaries.

Analysis of the shift toward inference-heavy AI workloads and implications for architecture, cost, and deployment strategies in the industry.

No substantive AI industry developments reported.

Opinion piece speculating that image generation capabilities represent progress toward AGI, referencing GPT-Image-2 adoption.

Applied Intuition deploys AI systems in mining, drones, trucks, and warships; CEO Qasar Younis and CTO Peter Ludwig discuss physical AI in adversarial environments.

DeepSeek releases V4 Pro (1.6T-A49B) and Flash (284B-A13B) models optimized for Huawei Ascend chips, no longer leading benchmarks.

Latent Space newsletter item referencing GPT 5.5 and OpenAI Codex Superapp with minimal detail; unclear if announcement or speculation.

Latent Space podcast covers AIE Europe conference insights and Agent Labs thesis on unsupervised learning and agent development trends for 2026.

Commentary on tokenization strategies as a recurring theme in AI industry discourse, without specific technical claims or announcements.

Shopify CTO discusses 2026 AI adoption roadmap, unlimited Claude Opus 4.6 token budget, and internal tools (Tangle, Tangent, SimGym).

OpenAI launches GPT-Image-2; Cursor secures $10B contract with xAI and $60B acquisition option.