Crypto
LDK’s Lightning preview adds flexibility—and migration homework
The v0.3-rc1 preview adds splice flexibility, while its compatibility notes give wallet developers homework before an upgrade.
Geeknewz series
The daily download
12 stories for curious minds.
AI
ShadowPEFT gives a big AI model a smaller travelling companion
PEFT 0.21.0 includes a trainable shadow network. Its detachable design is intriguing, but it comes with extra computation and clear limits.
Tech
iPhone Duo opens wide—and starts at CAD$2,999
Apple’s foldable iPhone reaches Canada in October. The bigger canvas comes with a substantial price and a later Pencil promise.
Crypto
Ethereum’s developers are opening the question box on September 16
The Ethereum Foundation’s Protocol team has a Reddit AMA scheduled for 14:00 UTC. Questions are already being collected.
Tech
Cloudflare’s quantum-ready DNS signatures need a bigger envelope
1.1.1.1 can now check ML-DSA-44 signatures. That is an early step toward a much wider DNS migration.
AI
IBM’s forecasting AI would like a look at your time series
A September 9 technical walkthrough details Granite’s downloadable forecasting model. The useful question is how it behaves on your data.
Crypto
Litecoin Core’s latest update is about the plumbing
The 0.21.5.8 maintenance release focuses on validation, relay and resource handling—not price predictions.
Tech
Linux 7.2.6 arrives with the unglamorous work that matters
Today’s stable kernel update is a useful reminder: maintenance releases deserve attention too.
AI
Google’s autofinetune puts AI experiments on the night shift
An agent can adjust training runs while you sleep. Someone still has to define what counts as an improvement.
Tech
Firefox is speeding up its release calendar. Your tabs can relax.
Mozilla’s two-week release rhythm is about getting finished work out sooner—not doubling the feature firehose.
AI
Google’s Kotlin agent toolkit gives local AI a seat at the table
The toolkit connects on-device and cloud models. Choosing where the work happens still matters.
AI
Your AI agent got an A. Did it actually do the homework?
A benchmark score tells one story. The checks your coding agent actually performs tell another.