Bytescope monitors hundreds of tech, eng and science publications for changes, using either traditional page comparison or vision capable LLMs. It uses llm (experimenting with fine-tuned llama3.2 8/70b) to verify if new content matches tech/engineering/science criteria.
I built this because keeping up with tech was eating hours of my day. I tried brutalist.report but needed keyword filtering and a wider range of monitored blogs.
I've been using it for a few days now and it seems to work ok. A bit slow but currently it all runs on a single server.
Any feedback is appreciated.
Edit: Forgot to mention, in case it's not clear - leave the input blank or use * for all updates (max 500 shown).
Cyberhaven suppose to be a security company
----
The attack began after a hacker successfully targeted a Cyberhaven employee via a phishing email that was sent to Chrome extension developers. The employee, believing the email was an official Google contact, clicked the email and input their login credentials on the phishing page.
I've developed an open-source inference server called Inferenceable with the aim of providing developers with a straightforward solution that can be installed without fussing over dependencies.
Inferenceable, built in Node.js, simplifies the process of building and running complex libraries like llama.cpp, making it accessible to developers of all levels.
It's also pluggable and can be used with your custom strategies.
There are 3 main functions
- General text based inference
- Image inference
- Generate text embeddings
Inferenceable also includes a simple UI which can be customised or deployed as is, pluggable authentication, CSP and rate limiter.
I'm working on adding more authentication strategies, including social logins, and creating additional examples.
One day this might be very useful in product design. It seems very close to creating a finished design… or Atleast a great starting point and a mental unblocker…
