Artificial Analysis v5 and the benchmarker's dilemma
Artificial Analysis v5 reviews benchmarking industry practices and gaps.
News Hub
What actually shipped in agent engineering, pulled from the labs, arXiv and Hacker News.
See who we follow24 articles
ResearchInterconnects·16d ago·3 min readSummaryListen 0:46
Podcast: Kimi K3.1 Pro, the new cyber reality, and Muse musingsInterconnects podcast episode 20 discussed Kimi K3.1 Pro and related topics.
ResearchInterconnects·20d ago·3 min readSummaryListen 0:44
Latest open artifacts (#25): DeepSeek V4.1 Flash 1017, American-made models, and new regulatory controls in the US & ChinaTitle alone insufficient for concrete summary.
Open sourceInterconnects·23d ago·3 min readSummary
Meta shakes up the frontiers, open and closedMeta releases open-weight models alongside closed models.
ResearchInterconnects·34d ago·3 min readSummary
The Cyber Risk Discourse is BrokenArticle discusses cyber risk discourse issues related to open-weights models and ideological trade-offs.
Open-Source AI & Open Models Reading List
Reading list compiled to help engineers understand open-source AI models and their implications.
ResearchInterconnects·62d ago·3 min readSummaryListen 0:45
When will average people feel AI’s impact?AI impact on average people remains uncertain despite five years into rapid technological advancement.
ResearchInterconnects·62d ago·3 min readSummaryListen 0:44
Latest open artifacts (#24): Motif-3, GLM-5.3, Hy4-preview and open model licensesNew open models Motif-3, GLM-5.3, and Hy4-preview released; open model licensing frameworks evolving.
ResearchInterconnects·84d ago·3 min readSummaryListen 0:41
Teaching Everyone to Fish for TokensNvidia encourages customers to build their own models rather than purchase from Anthropic or OpenAI.
ResearchInterconnects·87d ago·3 min readSummaryListen 0:55
GLM-5.3: How Chinese labs keep stride with the frontierChinese AI labs maintain competitive performance with frontier models through methods beyond distillation techniques.
ResearchInterconnects·91d ago·3 min readSummaryListen 0:49
5 useful things you'll learn in my new post-training textbook (shipping now!)A post-training textbook documenting lessons from training open models is now available for shipping.
IndustryInterconnects·98d ago·3 min readSummaryListen 0:39
Introducing our Artifacts Hub and Adoption DashboardInterconnects introduces an Artifacts Hub and Adoption Dashboard for curating and measuring the open ecosystem.
ResearchInterconnects·99d ago·3 min readSummaryListen 0:43
Latest open artifacts (#23): Laguna S2.1, Inkling, & Kimi K3 show the utility of open models on the Pareto frontierOpen models like Laguna S2.1, Inkling, and Kimi K3 demonstrate strong performance on the Pareto frontier.
ResearchInterconnects·110d ago·3 min read
Open models recap: more on Kimi K3, Qwen 3.8, Xi's WAIC speech, distillation, the open-closed gap, and what's nextPodcast discusses open models including Kimi K3, Qwen 3.8, distillation, and open-closed gaps.
ResearchInterconnects·112d ago·3 min read
Kimi K3: The open-weights escalationKimi K3 open-weights release has global implications for the AI ecosystem.
ResearchInterconnects·140d ago·3 min read
GLM-5.2 is the step change for open agentsGLM-5.2 represents a capability threshold for open source AI agents.
Open sourceInterconnects·143d ago·3 min read
Banning Open Source AI Would Be A MistakeBanning open source AI would be a mistake according to this op-ed.
ResearchInterconnects·148d ago·3 min read
Welcome to the AGI era of AI governanceArticle discusses governance challenges presented by advanced AI systems.
SafetyInterconnects·153d ago·3 min read
Claude Fable 5 and new AI safety fablesClaude Fable 5 and new AI safety fables released exploring power dynamics in frontier AI systems.
ResearchInterconnects·160d ago·3 min read
Farewell Ai2Author departed Allen Institute for AI after working on Olmo models.
The Agent Architect
One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.