18 sessions
PodcastThe Birth and Growth of Spark: An Open Source Success StoryClaimMatei Zaharia says Spark began as a PhD project focused on programming with large-scale data and was motivated by the limitations of existing tools for machine learning and interactive queries.9:13
Age of Industrialized AIClaimDan Jeffries says large thinking models will orchestrate tools and other models throughout software workflows, including game development.2:42
LangChain: Enabling LLMs to Use ToolsClaimLangChain has mainly focused on prototyping, but is starting to think about what it takes to run chains and agents in production.0:59
Generative Interfaces Beyond ChatClaimA conversation includes shared context, the speaker's intent, the recipient's interpretation or internal thinking, and an action.4:05
Agentic Relationship ManagementClaimHearth AI is developing agentic relationship management products that use AI agents to learn from, understand, and act on behalf of users.0:28
Reasoning MachinesClaimManually supplying information to an LLM does not scale, so tools can retrieve the necessary data from data stores.9:47
Guiding LLMs While Staying in the Driver's SeatClaimAdept AI is building a natural-language software collaborator that controls software tools such as Airbnb and Salesforce through mouse and keyboard actions.0:31
LLMs as Intelligent AssistantsPushed backSarah Aerni rejects replacing human review with fully autonomous generated output and argues that human review remains critical.23:40
Building Production CopilotsPushed backTristan Zajonc says autonomous agents do not work reliably yet, while the host expresses doubt based on what has been implemented so far.17:18
Building Reliable AI AgentsPushed backFully autonomous agents are not yet the right default for production; more deterministic, code-driven agents should be used instead.9:12
Guardrails for LLMs: A Practical ApproachClaimShreya Rajpal says a failed verification can trigger a new prompt containing context about what went wrong, so the model can try to correct its output.5:17
Using LLMs to Power Consumer Search at ScalePushed backAravind Srinivas rejects describing Perplexity as merely an LLM wrapper, arguing that the product requires substantial orchestration and software engineering around the models.18:18
PodcastBuilding Cody, an Open Source AI Coding AssistantPushed backDemetrios Brinkmann questioned whether low-code and no-code tools already addressed the problem, while Beyang Liu argued that their abstractions limit expressivity when requirements go beyond the platform's anticipated use cases.45:03
LLM SecurityPushed backThe idea that language models can soon operate autonomously in production was qualified by Uri Shamay, who said current models still fail at simple multi-step instructions.25:10
What Drives GenAI Development in the Next 3 YearsPushed backEuro Beinat says that reliable agents at scale are still difficult and suggests that near-term agents will focus on specific niches, while leaving room for the audience to disagree.10:12
The Truth About AI AgentsPushed backSilen Naihin disputes the idea that current AI agents are ready for reliable commercial use, saying that they often loop, become confused, or fail to complete tasks correctly.6:26
MeetupLLMs Mini Summit // MLOps Mini Summit Meetup #3ClaimThomas Capelle says local LLM assistants can combine speech recognition, language models, tool calls, and text-to-speech.5:32
PodcastBuilding the Future of AI in Software DevelopmentClaimCodeium uses off-the-shelf tools such as Prometheus and Grafana for general monitoring but builds core LLM systems itself when self-hosting and tight integration require it.50:14