Cool proactive agent experience: I shared my location with Instinct while traveling and asked it to flag points of interest, vistas, and good food spots near me as I walked around. Loc data is a bit delayed but other than that not bad! Also had it figure out public transit and directions. As a result I spent a lot less time fighting Google Maps with limited connectivity. Actually ended up looking at my phone a lot less

In the spirit of @__tinygrad__, we will be offering a $200 bounty for anyone who implements @Etched support in Tinygrad, along with a 100% Python Etched userspace driver. Must be <10k LoC to qualify.

Okay, I'm off to sleep, the cybersecurity eval suite build will continue, possibly all night. Crossed 440,000 loc so decided it was probably time to push something. Looking forward to waking up and possibly having the first 16 tasks fully completed so I can start testing, poking holes in them, and improving. And a bunch of ppl asked me about Fable. Yeah, probably won't benchmark Fable on this as it will hit way too many guardrails, but will test Opus 5, and of course every other model that should be decent at cybersecurity.

# Neo4j Features and Practical Usage ⚡ "Even in a graph DB, index design is 80% of performance." Put the right index on your MATCH anchor property and lookups go from O(n) to O(log n). 🏷️ Title: Range / Text / Point / Composite index 🔗 URL: 📘 Overview An index is a copy of nodes, relationships, or properties that provides a fast access path to the primary data. Once created, the DBMS keeps it updated automatically. It speeds up the anchor (starting point) of a MATCH, and the Cypher planner selects it automatically based on the predicate. ⚙️ How It Works ・Range (default): solves the most predicates—equality, range comparisons, `IN`, and `STARTS WITH`. This is the first index to reach for. ・Text: `STRING`-only, optimized for `CONTAINS` and `ENDS WITH`—ideal for substring search screens. ・Point: for spatial `POINT` values, optimized for distance queries and bounding-box searches. ・Composite: indexes several properties together, solving multi-condition filters in one go. ・Token lookup: a foundational index that speeds up label / relationship-type lookups. ・Beyond these, full-text indexes and vector indexes (similarity search / GenAI) are also available. 🛠️ Practical Usage Creating range / composite / text / point indexes on anchor properties: ```cypher CREATE INDEX user_email IF NOT EXISTS FOR (u:User) ON ( CREATE INDEX order_composite IF NOT EXISTS FOR (o:Order) ON (o.customerId, o.status); CREATE TEXT INDEX product_name_text IF NOT EXISTS FOR (p:Product) ON ( CREATE POINT INDEX store_loc IF NOT EXISTS FOR (s:Store) ON (s.location); ``` Inspect and drop them: ```cypher SHOW INDEXES; DROP INDEX user_email IF EXISTS; ``` 💡 Use Cases ・Apps that anchor MATCH on ` `Order.orderId`, or `Product.sku`, making lookups O(log n). ・Search screens that pick range for `STARTS WITH`, text for `CONTAINS`/`ENDS WITH`, and point for geo queries. ⚠️ Caveats ・Which index applies depends on the predicate. `CONTAINS` is not served by a range index—it needs a text index. ・Indexes add write cost and storage. Confirm they are actually used with `PROFILE` before relying on them. ・Composite indexes apply from the leading property onward, so column order affects performance. #Neo4j# #Cypher#