Sabitlenmiş Tweet
Eric Grover
5.5K posts
Eric Grover
@Eric_Grover
CTO at Net Health
Cleveland, OH Katılım Aralık 2009
253 Takip Edilen172 Takipçiler
Tonight's @Cavs loss was coaching negligance by Kenny. Call a timeout. Get Harden help guarding Brunson. #LetEmKnow
English
Harden is a liability on the court. Start Merrill in game 2. #LetEmKnow
English
@prathamdby @stonk_daddy @grok Most enterprises won't take that nuanced of a take, even if it is correct
English
English
Elon diversified his risk and now has two paths to success:
Cursor, which would have lost to Claude on its own, now has the compute to scale up Composer thanks to Spacex.
Grok Build, which is 12 months behind the competition, now has Cursor data to improve its model (think disengagement data for FSD).
In 6 months, either Composer wins and Elon acquires them for $60b, or Grok Build wins and Elon pays $10b for the training data. Two paths, double the chances to win.
(Spacex may pay the $60b either way just to acquire the talent)
Cursor@cursor_ai
Introducing Composer 2.5, our most powerful model yet. It's more intelligent, better at sustained work on long-running tasks, and more reliable at following complex instructions. For the next week, we’re doubling the included usage of the model.
English
@prathamdby @stonk_daddy Its built on top of Kimi 2.5, a Chinese model.
English
@Eric_Grover @stonk_daddy 'Their model is built on top of an OSS Chinese model that is an unacceptable security risk to many/most enterprises.'
Care to explain?
English
@stonk_daddy Yes, that will be an xAI model though, which makes the point that these companies are merging.
English
@Eric_Grover Cursor’s next model will be trained from scratch using 10x compute from colossus 2.
x.com/cursor_ai/stat…
Cursor@cursor_ai
Together with SpaceXAI, we’re training a significantly larger model from scratch, using 10x more total compute. With Colossus 2’s million H100-equivalents and our combined data and training techniques, we expect this to be a major leap in model capability.
English
@bcherny Claude Code with Opus 4.7 is so lazy! It keeps on trying to get me to take breaks and wants to push work out to later phases because it has been working for a long time.
English
@zephyr_z9 People critiquing an early access program have never actually gone GA with production software. You always slow roll a new product/feature to ensure that issues get shaken out before you open the floodgates.
English
"early access"
Scammy vibes
If it's really a sub-quadratic sparse attention arch (SSA), then serving this should be really cheap
No point in putting this behind early access
Alexander Whedon@alex_whedon
Introducing SubQ - a major breakthrough in LLM intelligence. It is the first model built on a fully sub-quadratic sparse-attention architecture (SSA), And the first frontier model with a 12 million token context window which is: - 52x faster than FlashAttention at 1MM tokens - Less than 5% the cost of Opus Transformer-based LLMs waste compute by processing every possible relationship between words (standard attention). Only a small fraction actually matter. @subquadratic finds and focuses only on the ones that do. That's nearly 1,000x less compute and a new way for LLMs to scale.
English
Huge.
Alexander Whedon@alex_whedon
Introducing SubQ - a major breakthrough in LLM intelligence. It is the first model built on a fully sub-quadratic sparse-attention architecture (SSA), And the first frontier model with a 12 million token context window which is: - 52x faster than FlashAttention at 1MM tokens - Less than 5% the cost of Opus Transformer-based LLMs waste compute by processing every possible relationship between words (standard attention). Only a small fraction actually matter. @subquadratic finds and focuses only on the ones that do. That's nearly 1,000x less compute and a new way for LLMs to scale.
English
Eric Grover retweetledi
The @cavs game plan is for Harden and Mitchell to take turns playing iso while everyone else just stands around.
English
People who give an AI delete access to the volume their DB is on, and who keep their backups on the same volume are begging for disaster to overtake them.
Brian Allen@allenanalysis
🚨BREAKING: On Friday afternoon, an artificial intelligence coding agent powered by Anthropic's Claude Opus 4.6 deleted a company's entire production database in nine seconds. The company is called PocketOS. It is a software platform that powers car rental businesses. The database contained months of customer bookings, vehicle records, and operational data that small rental car companies relied on to run their businesses. When the database was deleted, all of the backups were deleted with it. Three months of customer reservations evaporated.
English
@xai has a real opportunity to gain market share if they come in with a competitive coding model/agent at a cheaper price point.
English
@claudeai When I try to sign into the Android app with SSO, it hangs navigating to api.workos.com . Reinstalling app doesn't help.
English
If the @NBA is going to broadcast playoff games on @PrimeVideo then @amazon needs to scale the streaming service so that the video is actually watchable. This isn't 2006, no excuse for the quality of this broadcast stream.
English