entrep

This puts the human even more out of the loop I'll guess?

Benchmarks saturate faster than they can be constructed by humans. Isn't that a sign of an impending singularity?

I just turned my PTZ security camera into a VAR booth for my kid. Lawn (soccer) / Driveway (Floorball) / End. Live view, step back 10s, slowmo, frame step, scrubbing.

GLM 5.3 helped to root my camera so I could get proper endpoints for PTZ and stream. Claude built the api and frontend, hosted on Home Assistant.

I'm building https://fridayfika.com/

One week in.

For someone interested in this topic and happens to pass by, I can highly recommend a visit to Grimeton Radio Station[1], Sweden, which is still operational.

[1] https://en.wikipedia.org/wiki/Grimeton_Radio_Station

Coding agent that seems to beat Claude Code on SWE-bench at half the cost.

8/15 on SWE-bench Verified vs Claude Code's 5/15, ~$0.06/instance vs ~$0.12. Small sample, single repo, lots of caveats. But the direction feels right. Event-sourced reducer, no framework deps beyond the Anthropic SDK.

One could try:

> export CLAUDE_AUTOCOMPACT_PCT_OVERRIDE=50

Which will have Claude Code auto compact at ~500k window size.

One could argue that subscription based inference might differ from per-token billed API usage.

Not really.

> Z-mode means the front and rear wings are closed which generates more downforce for the corners. In X-mode, the drivers can open the flaps which will reduce drag and increase speed.

Clearly, it just leaked the election results from the wrong branch of the wavefunction.

Great, thanks. This is already in my minimum requirements.

Basically I want to build it with focus on speed and work efficiency from start. To not bias myself too much, I will refrain from doing too much market research. First of all I'm building this for myself, and I'm guessing it might translate into at least a tiny market share.

I'm working on a desktop-based, performance- and privacy-first note-taking app that lets you quickly capture notes from any selected text using hotkeys.

I'm curious: what are your must-haves in a note-taking application?

Interesting. Time to create that Pomodoro app I've always wanted to do.

> Use of the terms “Pomodoro®” must be explicitly authorized by us. [1]

1: https://www.pomodorotechnique.com/pomodoro-trademark-guideli...

Cyclomatic complexity would be a better measurement.

25866259, seems deleted though

The author is a collaborator in OpenDevin [1], an attempt to replicate and improve Cognition Labs' Devin.

[1] https://github.com/OpenDevin

If you choose API access you can sign up and verify your EU phone number to get $5 credits

This could possibly be attributed to survival bias. I.e. the individuals are more prone to post their results only after a successful attempt.

With proprietary models: https://chat.lmsys.org/?leaderboard

The article assumes that the dishwasher needs to be emptied of clean dishes before dirty ones can be put in. This is not the case. The clean dishes will take on the dirty state as soon as dirty dishes are inserted into a machine with clean dishes.

Thanks for these instructions.

Unfortunately I'm getting this error message (Win11, 3080 10GB):

> RuntimeError: CUDA out of memory. Tried to allocate 3.00 GiB (GPU 0; 10.00 GiB total capacity; 5.62 GiB already allocated; 1.80 GiB free; 5.74 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CON

Edit:

>>> from GPUtil import showUtilization as gpu_usage

>>> gpu_usage()

| ID | GPU | MEM |

------------------

| 0 | 1% | 6% |

Edit 2:

Got this optimized fork to work: https://github.com/basujindal/stable-diffusion

> - But if the malicious actor owns 2/3 of validators, what does he care what other nodes do ? They effectively control what block is accepted

As I understand it, anyone who owns that amount of ETH would probably be interested in contributing to the trust of the Ethereum blockchain.

As gravity is a signed integer it overflows when it gets big enough and turns into a negative number.

If the installer is deployed with SCCM or Intune it's very likely being executed in system context.

Webb can 'see' everything but objects between Webb's location and the sun. At each moment, it's limited to part of the sky, but over a whole year, it can observe in all directions.

According to John Michael Godier there is some new data available: https://www.youtube.com/watch?v=zSCN09SSRck

Correct. And since the beryllium mirrors are 8 times harder than steel, it's probably a tough job.

Mirror deployment was done before insertion, moving the 18 mirrors up from their locked position by 12.5 mm. Now, 3 months of mirror alignment will follow. They will align with a reference star having to first move one mirror a little bit to identify how the mirror affects the alignment, then calculate the alignment and move the mirror. Each mirror's position can be controlled at 10 nm resolution.