HydrallHarsh

Learning and Sharing | Open Source | Documeting

[HydrallHarsh @hxrhdev]

@vinncodes_ Thanks!

[HydrallHarsh @hxrhdev]

Have an interview tomorrow at a startup
🫡

Wish me luck!

[HydrallHarsh @hxrhdev]

Crazyyyy Lock in from Argentina

[HydrallHarsh @hxrhdev]

@cognee_ @WeMakeDevs It's live on PyPI 👇
pip install mithril-cognee
📦 https://t.co/eI7Ao86gmy
💻 https://t.co/rvrEKlpLrC

[HydrallHarsh @hxrhdev]

Shipped Mithril 🛡️ for the @cognee_ x @WeMakeDevs Hackathon - a security firewall for AI agent memory.
AI agents can now remember things long-term. Problem: they'll remember ANYTHING, including lies. One poisoned claim in shared memory = every future agent inherits it.

[HydrallHarsh @hxrhdev]

@cognee_ @WeMakeDevs Cognee handles the actual remembering + graph reasoning. Mithril just decides what's allowed to become "truth" in the first place.
Use it via Python SDK, REST API, or MCP server (so Claude Desktop / Cursor can call it directly instead of writing to memory raw).

[HydrallHarsh @hxrhdev]

Been writing a lot of code with AI lately…

Time to start writing code again before I depreciate 😭

[Vasilije @tricalt]

Introducing Cognee v1.0: a major breakthrough in agentic intelligence.

It is 145% better than Opus 4.8 and GPT 5.5 at long context memory retrieval.

Cognee allows a 100 BILLION token context window 100,000x more than Claude. It's:

- 6.9x cheaper than GPT 5.5 and Opus 4.8
- Cold starts in 350ms & searches in 260ms

Why this matters:

Today agents forget important context, redo tasks, waste tokens, and slow down as workflows get more complex.

Cognee solves this.

It’s not a place to build agents. It connects to the agents you’ve already built, across any platform, and makes them significantly cheaper, faster, and more accurate.

Here's how it works:

[HydrallHarsh @hxrhdev]

@connoratlunon https://t.co/X3uKlLHhTg

[HydrallHarsh @hxrhdev]

@cognee_ Sure!

[HydrallHarsh @hxrhdev]

Read through some @cognee_ docs today
A lot more to go through 👍

[HydrallHarsh @hxrhdev]

@cneuralnetwork Technoblade 🥲

[HydrallHarsh @hxrhdev]

Looks interesting 😀 @WeMakeDevs @cognee_
Probably will participate
Lemme know if anyone is down
https://t.co/Us9KECYdie

[HydrallHarsh @hxrhdev]

Really Decent Video on Agents
Probably will try implementing a Agent loop from scratch - Will post if I write one

[HydrallHarsh @hxrhdev]

Was making a project around News Aggregation - for AI specifically

Found this really detailed blog by @0xkishankumar - https://t.co/bDsIW8sfy0

Really good read!

[HydrallHarsh @hxrhdev]

@WeMakeDevs @WithCoral_com It uses Coral to investigate GitHub changes, vulnerable dependencies, Notion policies, Slack context, and possible secret leaks then turns it all into a clear risk report.

[HydrallHarsh @hxrhdev]

Built HarborGuard for the @WeMakeDevs @WithCoral_com hackathon ⚓️

It’s a security investigation dashboard for that awkward moment before someone says “ship it.”

https://t.co/XGZKsomh9n

[HydrallHarsh @hxrhdev]

@WeMakeDevs @WithCoral_com :)