@useTRMNL@remarkablepaper Neat! Got it working with KOReader. I tried the linux approach on the BYOD page but that didn't seem compatible with the reMarkable's unique framebuffer. So far so good with KOReader + trmnl-koreader though.
Adding fog of war to https://t.co/Ordcxg6eIw - clusters of three or more drawings clear it, so neighbourhoods are easy to spot but lonesome tiles stay hidden. Hovering over the fog lets you peek through and select the tiles under it.
Really hoping the attackers donโt leak my repos full of:
- six competing notes app prototypes
- readmes that just sayโTODOโ
- projects with 83 dependencies and one feature
Please respect my privacy at this difficult time.
๐จData Breach Alert โผ๏ธ
๐ง๐ฒ๐ฎ๐บ๐ฃ๐๐ฃ ๐๐น๐ฎ๐ถ๐บ๐ ๐ฆ๐ฎ๐น๐ฒ ๐ผ๐ณ ๐๐ถ๐๐๐๐ฏ ๐๐ป๐๐ฒ๐ฟ๐ป๐ฎ๐น ๐ฆ๐ผ๐๐ฟ๐ฐ๐ฒ ๐๐ผ๐ฑ๐ฒ
TeamPCP hacking group claimed the compromise and sale of GitHub internal data, allegedly including around 4,000 private repositories containing source code related to GitHubโs main platform and internal organizations.
Threat actor: TeamPCP
Sector: ICT
Data exposure (claimed): Approximately 4,000 private repositories
Data type: Source code
Observed: May 19, 2026
Status: Pending verification
ESIXยฉ: 7.96
Full details and impact assessment on https://t.co/eB7qgxKFAa
@webguy@GergelyOrosz Yeah, had a play with implementing my own client (AI-assisted). All I was using was one endpoint on the Responses API. No streaming, all very basic. Shaved 30% off my app's binary size! https://t.co/5X25kppsGB
Just tried moving off OpenAI's Go SDK in favour of a small client covering just the Responses API... it reduced my binary's size by 30%!
https://t.co/SSx4G9j1r6
Just tried moving off OpenAI's Go SDK in favour of a small client covering just the Responses API... it reduced my binary's size by 30%!
https://t.co/SSx4G9j1r6
@GergelyOrosz If this ends up with OpenAI moving their Go SDK off Stainless, that's a win in my book. The thing currently adds 11MB to my project's binary... nearly half the size of ALL my deps combined.
@thdxr Been experimenting with this, what works for me is annotating the generated code with "TODO LLM: x" comments, then prompting the model to grep and work through them. (compacting between feedback rounds if need be)
GNU Parallel book was very much worth the read. Helped me turn a ~14 line bash script into one command :)
Wrote a post with more details: https://t.co/MjQhPI2BNl
Wow, reading through the GNU Parallel book and it's hilarious right from its opening dedication...
(it's also free under creative commons, here: https://t.co/fuhKG1S8km)