BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Palaner//palaner Single Event//EN
METHOD:PUBLISH
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:luma-08351970c0@palaner.app
DTSTAMP:20260802T150632Z
DTSTART:20260807T130000Z
DTEND:20260807T143000Z
SUMMARY:Lossfunk Research Showcase
DESCRIPTION:Saved in Palaner as: saved\n\nPalaner match: 25% · why: matches
  your usual pattern\n\nLossfunk is a research lab based in Bangalore\, Ind
 ia\, focused on exploring general intelligence in all its glory.\nRead our
  charter HERE\nWe are putting together a research showcase for the evening
  where Lossfunk researchers will talk about various research projects they
  are working on.\nThis is perfect for people working across the wide spect
 rum of research in frontier intelligence science.\nJoin us for a curated r
 oom of fellow researchers.\nHere's a snippet of who will be presenting\nSu
 shrut Thorat will talk about what we are cooking at Lossfunk. He will trac
 e our research since inception\, share where we’re headed\, and conclude w
 ith a glimpse into an esoteric project exploring the nature of consciousne
 ss.\nSukanth K will introduce his fascinations with the theme of "transhum
 anism\," tracing his early work on cognitive augmentation using AI and sma
 rt glasses and his deeper shift into sensory augmentation\, exploring whet
 her AI can become a new sense. He will also outline the role transhumanism
  plays in a post-Singularity world\, and how his current research at lossf
 unk contributes to that future.\nDhruv Trehan and Abhinav Kumar will share
  findings from their work on LLMs for Science\, including documenting why 
 LLMs are not scientists yet\, measuring research taste\, studying attracto
 rs in autoresearch systems\, using evolutionary search such as AlphaEvolve
  to discover RL interfaces\, and building an agentic research mentor. Toge
 ther\, these projects reveal both the potential of LLMs in research and th
 e biases and failure modes that shape how they explore\, judge\, and condu
 ct science.\nAyush Nangia will talk about his work on context compaction i
 n agents. Every agent that runs long enough has to compress its own memory
 \, and almost everyone judges that compression by how much information sur
 vived it. But he thinks that is the wrong question\, and he is measuring a
  different one: whether the agent still does the same thing after you comp
 ress its history.\nWe have a limited number of spots so we would encourage
  you to register swiftly.\nThis event is part of Basecamp\, a community-po
 wered tech week in Bengaluru bringing founders\, builders\, and operators 
 together. Basecamp is free for the community and powered by the community.
URL:https://luma.com/2dza1f3z
CATEGORIES:palaner,saved
STATUS:TENTATIVE
END:VEVENT
END:VCALENDAR