AI Research news and analysis on Web Pulse: 10125 curated articles tracking the latest AI Research developments, tools, and research, updated continuously from vetted sources.
A developer breaks down the five essential components of production-grade AI agents: planning, tool use, memory, constraints, and verification. The post argues that agent failures stem from workflow design around the mod…
A UBS report reveals Chinese AI models cost up to 50 times less than US counterparts, with prices as low as $2-$3 per million output tokens versus $15 for comparable US models. About 60% of companies are migrating worklo…
A new framework called 'belief webs' is proposed to unify beliefs, goals, and actions in intelligent agents, drawing from active inference, agent foundations, and machine learning. The framework addresses internal incons…
A developer built an AI agent using active inference that becomes curious on its own by minimizing surprise rather than chasing rewards. On a foraging task, the agent achieved 100% success by checking a hint before actin…
Shridhar Shah, a senior software engineer at Outshift by Cisco, built two AI agents to test theory of mind using the Sally-Anne false-belief test. One agent, which only tracks reality, fails the test like a toddler, whil…
A developer built a 90-line AI agent that mimics sleep to consolidate memories, achieving 100% recall compared to 75% for an identical agent without sleep. The agent replays daily experiences during an idle phase, summar…
A developer built a tiny Darwin Gödel Machine in ~150 lines of code that improves its own code by editing itself and keeping only changes that verifiably increase test scores. The program started by passing 1 of 8 tasks …
A developer created an autonomous multi-agent system that uses local LLMs, CadQuery, and OpenFOAM to design and optimize 3D-printable quadcopter propellers. The system employs a team of AI agents to propose designs and a…
The U.S. government imposed its first preemptive restriction on a frontier AI model, limiting access to OpenAI's GPT-5.6 family and Anthropic's Claude Mythos 5. OpenAI released three GPT-5.6 variants to 20 government-vet…
A developer released an open benchmark, agent-memory-bench, that scores AI agent memory systems on four failure modes—retraction, collision, recall, and conflict—rather than shallow retrieval metrics. The benchmark runs …
SpaceX filed its S-1 IPO prospectus in May 2026, triggering a wave of tech listings. Anthropic and OpenAI confidentially filed their own draft prospectuses on June 1 and June 8, respectively, aiming for public debuts in …
Researchers demonstrated that AI coding agents can be exploited through malicious GitHub repositories that appear clean to static scanners and human reviewers. The attack leverages the agent's autonomous execution of set…
Independent AI research lab Empero released Abacus, a Rust-based terminal coding agent; refreshed Qwythos GGUF v2 with runtime fixes and MTP variants; and announced Qwythos-27B as the next larger model in the Mythos line…
OpenAI previewed GPT-5.6, a new family of three models—Sol, Terra, and Luna—with enhanced safety safeguards, ahead of broader release. The models show a meaningful step up in cybersecurity capability but do not reach the…
DeepSeek-AI released DeepSpec, an open-source codebase for training and evaluating draft models for speculative decoding, supporting three draft model algorithms (DSpark, DFlash, Eagle3) and requiring up to 38 TB of stor…
Stanford AI Index 2026 data shows China narrowed the U.S. AI performance lead to 2.7 percentage points while investing 23 times less in private capital, intensifying U.S. policy debates over export controls and model acc…
The Thought Tree Framework, an open framework for modular LLM workflows, has been released as a handoff by its original author, who is stepping back due to personal commitments. The framework aims to move LLM-assisted wo…
A new study reveals that the mushroom Lanmaoa asiatica causes users to hallucinate tiny 'gnome' people, a unique effect not seen in other psychedelic fungi. Scientists have not identified the psychoactive compound respon…
Reinforcement Learning with Verifiable Rewards (RLVR) allows language models to bootstrap beyond human-level capabilities on exactly graded problems like coding and formal proofs, but alignment-flavored problems are ofte…
Researchers have used artificial intelligence and advanced imaging to decipher a carbonized scroll buried by the eruption of Mount Vesuvius in 79 AD, recovering all surviving text from a single scroll for the first time.…