cognee

This Thursday in London: the Grok Bot Community Engineering Hackathon, and cognee is one of the technology partners. It's an evening for community engineers, DevRels, organisers and builders. Teams pick a real community problem, like finding new members, running events or keeping people active, and have a working demo by code freeze. Organised by Francisco T., Louis L. and Keanu Czirjak. Venue partner • Expedite, made possible by William Schulz Technology partners • Supabase (Isabelle Roser) • cognee (Veljko Kovac) • Zero (Mariia Nedelkina) • SpaceXAI Community (Francisco T.) Community partners: • Unicorn Mafia (Charlie Cheesman) • basedcollective (Fatema Al Khalifa) • Kickstart Global (Ilya Vencjuns) •Hackradar (Keanu) • AI Agents Lab (May Yu & Minghan Xiao) • Corgi (Claudia Maggi) • Night Office (Oli Kingshott & Shreyas Raghu) • The Hack Collective (Louis) • United Italian Societies (Umberto Belluzzo) • Outpost (Andy Tyler & Oscar Falemara) • AI Pilled (Nuan Zhang & Yiran Qin) • Other (Zakee Abdi, Fergus McKenzie-Wilson) Thursday 22 October, 18:00 to 23:00 BST Expedite London HQ, 7 Lyric Square Free, host approval required, last spots open Apply: https://luma.com/eveur09a

Have a look at our hackathon last week with ScaleKit in San Francisco This is the infamous one, where Vas met the guy accused for spying at OpenAI

Four corners in the ring this Saturday: claude-mem, HelixDB, cognee, and the memory setup you bring. Mem-Off is an open comparison of AI memory systems in San Francisco, co-hosted with HelixDB and CMEM.ai, the team behind claude-mem. Every system runs the same scenarios, then goes through the pressure test: facts that go stale, updates, contradicting information, forgetting. Results go up live, with the people who built these systems in the room. Bring something you built yourself or a system you sell. Saturday, October 17, 10:30am to 6:30pm PT Pebblebed, San Francisco Free Bring your memory. Let's see what sticks: luma.com/cognee-cy86

This Thursday at 17:00 CEST, Luca Stromann and Veljko Kovac from our team build a company brain live with n8nn Lab. Free to join, and everyone who registers gets the recording.

Join us on Saturday, October 17 in San Francisco for Mem Off, the AI memory showdown we're co-hosting with HelixDB and claude-mem (CMEM.ai). Everyone claims their AI has memory. Mem Off is where we check whether it actually works. Bring your own memory setup, whether you built it yourself or sell it. Every system runs on the same scenarios and evaluation criteria, then goes through the pressure test: facts that go stale, updates, contradicting information, forgetting. Results go up live, with the people who built these systems in the room. 10:30am to 6:30pm PT, Mission District. Free to attend. Bring your memory. Let's see what sticks: luma.com/cognee-cy86

This Thursday at 17:00 CEST, Luca Stromann and Veljko Kovac from our team build a company brain live with n8nn Lab. Free to join, and everyone who registers gets the recording.

The cognee team will be at In The Loop by n8n in Berlin on Tuesday and Wednesday. Come say hi.

We’re partnering with Briefcase and Stripe on this amazing hackathon in London next Thursday. Theme is: AI agents in fintech. Three hours to build an agent that handles real financial operations. Judges are: • Ján Stehlík (Briefcase) • Fionn O Colmain (Stripe) • David Gelberg (10 Downing Street AI Team) • Dominic Hudson (AI at Meta) • Akash Bajwa (Earlybird Venture Capital) Thursday 15 October, 17:15 to 22:00 BST, Central London. Win a price of £2,000. Free, apply here: https://luma.com/3k1pazop

Long context is valuable. It helps when the relevant information can be passed directly into the model. But context still has limits. It disappears unless something stores it. Long prompts increase cost and latency. Models do not use long context perfectly. Long-running tasks eventually produce more information than one request should carry. Memory handles a different job: deciding what should persist between calls, sessions, agents, and tools. The context window is just the delivery mechanism.

Two systems can get the same final-answer score while maintaining very different memory stores. One may have clean entities, scoped facts, provenance, and update rules. Another may have duplicate customers, unresolved contradictions, stale preferences, and broken links that happen not to show up in today's eval. Production memory needs evaluation at multiple levels. Did the agent answer or act correctly? Is the memory state itself healthy enough to keep using tomorrow?