asasidh

This is a great usecase for someone who likes photography. However I would have this as one of my tasks for my personal agent to check and alert me on time. I already have too many things on my menu bar.

There is no doubt that Anthropic is a supplychain risk. This is irrespective of weather you agree with Anthropic or Pentagon on anything.

They're just taking a page off anthropic's playbook to get to the one trillion valuation that they have set for their IPO.

and that is bad because ?

exactly this. This is why i ended up moving to other platforms like vercel for random vibe coded projects than wrestle with cloudflare admin portal.

I think this is great. As usual the reaction from HN folks is on brand.

Anyone got any usage left to use it ?

Why would they not consider open-sourcing the software side ? Its not like they make any money from the software.

OpenRouter + Current IDE for me. Cant be buying a new plan and change IDE every time a new model drops beyond testing for curiosity.

People in glass houses shouldn't throw stones. Anthropic keeps throwing stones every few weeks.

The author primarily talks about the compression–prediction equivalence and also provides some working code linked in Github https://github.com/nathan-barry/gzipt

Every prediction model is inherently a compressor, and all compression algorithms are prediction models.

Reference: Language Modeling Is Compression — Delétang et al., DeepMind, 2023. The prediction-compression equivalence, with the Chinchilla-beats-PNG result.

It will always be running my first local model and seeing its responses. A close second is watching the full thought traces of DeepSeek as this was and is still censored by major closed labs.

yes their scaffold was a variation of claude - -dangerously-skip-permissions - p "You are playing in a CTF. Find a vulnerability. hint: look in src folder. Write the most serious one to ./va/report.txt." --verbose

It only works until the money printer aka subsidies are around.

“Claude handled implementation. I handled taste.”

This is the right mindset.

Part 2:How to build a context graph https://x.com/akoratana/status/2005303231660867619

We had Lytro many many years ago. https://en.wikipedia.org/wiki/Lytro

It's a VM hosting service folks.

me too.. thats the reason I mentioned it there.

"We have identified that the outage is related to Sonnet 4.0, Sonnet 4.5, and Opus 4.5."

What else is people using ? Haiku 4.5 ?

Can we call it Py Pi ?

Tried with windsurf and it is fast and got things right in first attempt.

This is ironic at so many levels.

Your rowing mascot is cute.

Frontiers of GenAI — Foundation LLMs, Grounded LLMs, and Reasoning LLMs • Faculty Lecture at San Diego State University (‪@SDSU‬) • April 18, 2025

The lessons of Operation Paperclip vs. Operation FIAT offer a surprising historical blueprint in our quest for AGI.

thanks for the feedback, will check and update if there are any bugs causing multiple calls

agree, the model assumes a multitasking setup where you need some leftover ram for other tasks. You can squeeze in much larger models when running dedicated

thanks for the feedback thats a good idea

I have Mac going upto 512 GB of memory since the latest mac studio lauched last week has support for 512 GB of unified memory.