Caura (@CauraAI) header
Caura (@CauraAI) profile picture

Caura

@CauraAI

Building the memory layer for the hyper-agent generation. Governed shared memory for AI agent fleets. Open source and https://nitter.gq/t.co/z92PMs8e9a

 Global
caura.ai
Joined September 2025
  • Tweets543
  • Following314
  • Followers1.9K
We gave a Databricks Agent Bricks agent and a Claude agent one shared memory over MCP. Told Databricks about a trip. Claude planned dinner around it, then corrected the date. A brand-new Databricks session already knew. ~530 lines. Full code 👇 caura.ai/use-cases/databrick…
Image from Caura (@CauraAI)
3
2
342
Meta Muse says it plays with the big guys. The big guys aren't impressed. 8 frontier models graded each other blind, 11,137 times. Muse Wonder debuted 6th. Claude Opus 5.5 pulled away: a 126-point Elo lead, up from 53 last week. caura.ai/blog/meta-muse-debu…
Image from Caura (@CauraAI)
1
3
427
We had 6 frontier models grade each other's work 3,570 times. They inflated their own scores (GPT-6 Astra: +0.76). They gave famous names a bump (up to +0.25). They favored whoever answered first. And the harshest judge finished #1. AIs have egos → caura.ai/blog/ais-have-egos/
Image from Caura (@CauraAI)
1
1
3
442
We counted every tool grant in a 5-agent fleet: 14 of 60 possible agent–tool pairs. Prompt-only enforcement: 0. That's the whole lesson of multi-agent coordination. A decision one agent makes has to become a constraint the next agent can't route around. caura.ai/blog/multi-agent-sy…
Image from Caura (@CauraAI)
1
2
7
477
Harness Engineering: Stop Asking Your Agents to Decide What to Remember medium.com/caura-ai/harness-…
Image from Caura (@CauraAI)
1
3
11
810
Time to Stop Asking Your Agents to Decide What to Remember Deterministic memory: why the harness — not the model — should own every read and write. With a 130-line cross-vendor demo: a Gemini agent writes, a Claude agent recalls. caura.ai/blog/harness-engine…
Image from Caura (@CauraAI)
2
2
8
1,022
One of your agents figures something out at 2am. By morning it's gone — and the next agent repeats the same mistake, burns the same three hours, and nobody notices. That's not a memory problem. It's a teamwork problem. It starts with one memory that survives the night — 5 minutes: memclaw.net/docs/tutorials/f…
Two panels: 2:04 AM, agent solves a problem — 'Found it,' solved. 9:00 AM, another agent re-solving the same problem — 'Found it…?', solving it again. Caption: one memory that survives the night.

ALT Two panels: 2:04 AM, agent solves a problem — 'Found it,' solved. 9:00 AM, another agent re-solving the same problem — 'Found it…?', solving it again. Caption: one memory that survives the night.

3
3
5
724
Replying to @jack
The engram is the underrated part of this. Private, encrypted, per-agent, survives restarts. The tier after it is shared: what still holds, who vouched, who may read. Wrote up four ways to compose the two — none built yet: memclaw.net/blog/buzz-agents…
Three stacked layers. THE LOG — everything that happened, signed, replayable, hash-chained — Buzz. THE ENGRAM — what each agent knows, encrypted under its own key — Buzz. SHARED, GOVERNED — what currently holds, who vouched, who may read — MemClaw. Headline: Buzz solved identity. The third tier is shared memory.

ALT Three stacked layers. THE LOG — everything that happened, signed, replayable, hash-chained — Buzz. THE ENGRAM — what each agent knows, encrypted under its own key — Buzz. SHARED, GOVERNED — what currently holds, who vouched, who may read — MemClaw. Headline: Buzz solved identity. The third tier is shared memory.

24
Block open-sourced Buzz — Apache-2.0. Every participant, human or agent, holds their own cryptographic keypair. Every action is a signed event. Every agent carries an encrypted memory. Identity was the hard part. @blocks did it first. What a fleet needs beside it: memclaw.net/blog/buzz-agents…
Three stacked layers. THE LOG — everything that happened, signed, replayable, hash-chained — Buzz. THE ENGRAM — what each agent knows, encrypted under its own key — Buzz. SHARED, GOVERNED — what currently holds, who vouched, who may read — MemClaw. Headline: Buzz solved identity. The third tier is shared memory.

ALT Three stacked layers. THE LOG — everything that happened, signed, replayable, hash-chained — Buzz. THE ENGRAM — what each agent knows, encrypted under its own key — Buzz. SHARED, GOVERNED — what currently holds, who vouched, who may read — MemClaw. Headline: Buzz solved identity. The third tier is shared memory.

2
4
1
11
1,112
Five frontier models, 100 questions, 2,922 blind pairwise matches. Claude Opus 5 wins at 8.87 — clearest result the protocol has produced. Claude Fable 5? Third. Again not beaten — it forfeited. Four blank answers, two of them high-school biology. A model that reasons at 9.29 finished LAST in factual knowledge, below Grok. Opus won. Fable forfeited → memclaw.net/blog/opus-5-won-…
Bar chart of PeerRank July25 final rankings: claude-opus-5 first at 8.87, kimi-k3 8.17, claude-fable-5 8.11, gpt-5.6 7.98, grok-4.5 6.64.

ALT Bar chart of PeerRank July25 final rankings: claude-opus-5 first at 8.87, kimi-k3 8.17, claude-fable-5 8.11, gpt-5.6 7.98, grok-4.5 6.64.

2
3
1
15
20,677
New agent joins the standup. "I've taken the liberty of rebuilding the reconciliation service." Team: "…we killed that in April." Agent: "I was not told about April." Nothing ever tells it about April. That’s the default. Every restart, every new agent in the fleet pays the same tax. memclaw.net/?utm_source=x&ut…
Image from Caura (@CauraAI)
2
3
8
816
Load more