🚀 Weekly Tech Recap: Cache Me If You Can (May 5-11, 2025)

Caching made even harder. Reddis goes back to Open Source & AI is already reprogramming humans!

Waiting for the Cache Puns to Expire.

Cache Hit or Cache Miss? There is no conceivable third option… or is there? The recently published Analysis of Similarity Caching on General Cache Networks introduces a new option to thoroughly confuse the matter - a “Similarity Hit”.

This essentially means the system found something “good enough” to be considered a hit, even though it’s not the exact thing you expected.

Good Enough means it found something with a high enough cosine similarity or small enough euclidean distance.

Caching is already a hard enough problem with those two options. Do we really need a third?

My first thought is - How could an attacker use this to publicly cache a resource carrying an exploit? Perhaps another paper is required to research that.

Reddis Open Source Again!

Keeping the cache tropes warm.

Back in March 2024 Redis was relicensed starting with version 7.4. It was announced that the core Redis software would transition from the BSD license to a dual-license, losing its OS status and forcing vendors to divest from free-use of their product.

The community immediately responded with the Valkey fork which has seen continuing success and support from major players.

On May 1, 2025 Redis 8.0, reintroducing an OSI-approved open source license option: the GNU Affero General Public License v3 (AGPLv3). This move restored Redis's status as an open source project.

Is this the same old story of cloud vendors abusing the goodwill of open source projects we’ve seen play out a many times already with Java, ElasticSearch, Terraform…?

ChatGPT Already Reprogramming Humans?

I must admit I expected this one to take a few months longer than it did.

"Chatgpt-induced psychosis." is now a phrase you can add to your outdated analog meat-based neural network.

According to a report from Rolling Stone, Reddit users are sharing stories of loved ones developing delusions through conversations with AI. Some cases involve users believing they've uncovered cosmic truths, been chosen for divine missions, or that the AI itself is sentient or godlike.

I suppose the strict definition of God is an omnipotent, omnipresent, infinitely intelligent being. So maybe this checks out?

OpenAI are yet to confirm if the ghost is actually in the machine.