AI News

Today's Highlights

Anthropic says Claude now writes 80% of its codebase, boosting engineer output eightfold as capabilities double every four months

#20

+36 more

Recent Stars

pedropark99/zig-book6H AGO
Provides the source and examples for a project-based introductory book on Zig.
1095
2.6k stars
seanmor5/honeycomb6H AGO
Performs fast LLM inference in Elixir using Bumblebee and EXLA.
862
69 stars
ethanhe42/nanoRL7H AGO
Implements minimal single-file SFT, DPO, GRPO, and PPO for language model fine-tuning on toy arithmetic tasks.
862
1692
108 stars
…tinual-learning-bench7H AGO
Benchmarks AI agents on learning from repeated interactions across multi-episode tasks with adaptation metrics.
862
846
127 stars
trycua/cua7H AGO
Supplies sandboxes, SDKs, and benchmarks for AI agents to control full desktops on macOS, Linux, and Windows.
862
773
17.6k stars

Github Stars

(7 days)

Today's Highlights

Anthropic says Claude now writes 80% of its codebase, boosting engineer output eightfold as capabilities double every four months

Top Stories

Recent Stars

Github Stars

AI News

Today's Highlights

Anthropic says Claude now writes 80% of its codebase, boosting engineer output eightfold as capabilities double every four months

Top Stories

Recent Stars

Github Stars

Yesterday's Top Stories, Jun 4, 2026.

Anthropic says Claude now writes 80% of its codebase, boosting engineer output eightfold as capabilities double every four monthsMythos Preview achieved a 52x code optimization speedup.

NVIDIA releases Nemotron 3 Ultra, a 550B parameter open-weight hybrid Mamba2-Transformer MoE model for agentic workloadsIt reduces operational costs by up to 30 percent.

Anthropic reports its engineers now ship eight times more code per quarter, prompting debate over how to measure AI productivityAndy Jones called the increase an accelerated technological takeoff.

Google Magenta releases Magenta RealTime 2, an open-weights model for local real-time music generation with sub-200ms latencyThe release includes companion apps and DAW plugins

OpenAI launches upgraded ChatGPT memory system using a "dreaming" process to carry context across conversationsFactual recall success rates rose from 41.5% to 82.8%.

SpaceX promotional campaign teases plans to build critical infrastructure across space, connectivity, and AIThe campaign video includes a link to spacexipo.com.

AI and biotech leaders, including Sam Altman and Demis Hassabis, urge Congress to mandate DNA synthesis screeningThe proposed laws would require customer identity verification.

Anthropic embeds six engineers at the NSA to deploy its Mythos AI system for offensive cyber operationsThe Financial Times first reported the specialized defense partnership.

LLM evaluation platform Arena launches Agent Mode to benchmark GPT-5.5, Claude Opus 4.7, and Gemini 3.1 Pro on multi-step tasksThe platform measures task success, steerability, and tool hallucination.

Cognition AI launches a $10 million productivity guarantee to refund enterprise customers if its Devin AI agent fails to deliver valueProductivity is calculated in engineering hours instead of tokens.

Goldman Sachs forecasts SpaceX AI revenue will hit $322 billion by 2030, drawing skepticism over its IPO underwriting roleCompute-as-a-service offerings are expected to drive the growth.

Ramp raises $750 million at a $44 billion valuation and launches tools to monitor enterprise AI token consumptionCEO Eric Glyman calls tokens the fastest-growing business cost.

UC Berkeley introductory CS failure rates spike to 35%, prompting 1,300 faculty to demand SAT and ACT reinstatementInstructors blame AI overreliance and weak foundational math skills.

Sam Altman, Dario Amodei, and Demis Hassabis urge Congress to mandate screening of synthetic nucleic acids to prevent AI biosecurity risksThe proposed rules also target physical DNA manufacturing equipment.

AI robotics company 1X launches World Model Lab led by former Luma AI founding engineer Samarth Sinha to train humanoid robotsIt avoids VLA wrappers by pretraining models on physics.

Anthropic proposes global verification systems to temporarily pause frontier AI development, drawing criticism from safety commentator Michaël TrazziAnthropic compared the coordination to historic nuclear disarmament treaties.

Google releases Gemma 4 12B, an encoder-free multimodal model under Apache 2.0 that beats the larger Gemma 3 27BThe model runs locally on laptops with 16GB VRAM.

OpenAI says its AI model found a counterexample to an 80-year-old mathematical conjecture by Paul ErdősReinforcement learning guided the model's search of mathematical spaces.

Raindrop AI co-founder Ben Hylak launches Raindrop 2.0, introducing self-healing agents to autonomously detect and triage agent workflow failuresCustom-trained models flag issues like looping build failures

US officials weigh taking equity stakes in AI companies following a proposal by OpenAI CEO Sam AltmanThe acquired shares would fund public dividends.

Policy scholar Dean W. Ball argues that fully autonomous corporations may require a global banThe debate was sparked by Argentina's national AI strategy.

George Mason University economist Tyler Cowen predicts AI integration will drive 2.5% annual GDP growthThe talk also details AI's impact on healthcare.

Sci-fi author Ted Chiang argues that artificial intelligence systems are not consciousThe essay prompted debate over Chiang's recent AI commentary.

Poke launches the first AI agent with official Apple approval to send text messages directly within iMessageIt integrates with third-party apps like Gmail and Notion.

Proposed federal Obernolte-Trahan framework would preempt state AI laws for three years, targeting models over 10^26 FLOPsIt codifies the Center for AI Standards and Innovation

Researcher Alex Imas argues high uncertainty makes scenario modeling better than timeline forecasts for predicting AGI's economic impactThe debate focused on taxing and redistributing AGI-generated wealth

Marc Andreessen claims Biden administration officials warned against founding AI startups, prompting his endorsement of Donald TrumpDaniel Eth argued Andreessen misinterpreted a market consolidation prediction.

Investor Slater Stich interviews @polynoamial on applying AI to the Erdős unit distance conjecture and IMO benchmarksThe video launches a new series on AI in mathematics

Anthropic pauses its red-teaming program after an unreleased Claude Oceanus model checkpoint leaks onlineThe leak reportedly originated from distributed red-team checkpoints.

Anthropic says its engineers ship eight times more code per quarter after deploying Claude CodeSmol.ai's Shawn Wang tracked rising Claude Code session durations.