Prashant Deva (@pdeva) profile picture

Geek. I like video games and math. Chief guy @ https://nitter.gq/t.co/ZjunoadPpL

San Francisco
Joined April 2007
Photo by Prashant Deva (@pdeva)Photo by Prashant Deva (@pdeva)Photo by Prashant Deva (@pdeva)Photo by Prashant Deva (@pdeva)Photo by Prashant Deva (@pdeva)Photo by Prashant Deva (@pdeva)Photo by Prashant Deva (@pdeva)Photo by Prashant Deva (@pdeva)Photo by Prashant Deva (@pdeva)Photo by Prashant Deva (@pdeva)
Apparently OpenAI now blocks Claude from accessing all openai domain pages... Asked Claude to compare OpenAI's approach with Cursor and Anthropic VCS and Claude cannot read it.
We built Pando, a new Git backend for our monorepo, to handle more pushes and fetches without changing how our engineers and agents use Git. Pando cut push latency by ~80% and now serves nearly a petabyte of Git data a day. We can grow Git capacity alongside agent usage.
1
147
at this point agent sandbox startups are being spun up faster than sandboxes..
1
2
107
me: Opus make me a cache for these files Opus: here is a cache. but i introduced a new intermediary format, that keeps only certain fields in compressed serialized form in memory and deserializes on demand. I measured the perfect ratio of cpu/mem balance to keep only the right fields serialized. me: 🤯
86
give Astra computer use access to Photoshop and tell it to make your favorite photo look like a frame from a chosen movie.. the results will be insane… it will even crop and fix the aspect ratio to match the movie. 🍿
59
Sol in Ultrafast mode comes to screeching halt during the constant 'Compacting Context'.. like putting speed breakers in front of your F1 car. OpenAI is the only major provider still stuck at sub 1M context...
1
74
Me: I want to parallelize this task completely dynamically based on changing workload. Opus: Here is a trick from TCP windowing... Me: 🤯🤯
1
1
2
169
new Anthropic version control seems closer to Perforce than git. What’s old is new again.
New on our Engineering blog: Creating in-house source control. In May, we merged around 2,500 code changes per day. In September, we merged more than 30,000 per day. As a result, we became bottlenecked by developer infrastructure built for human (rather than agent) scale. Here, we share how we built a new version control system—and fully migrated our monorepo from git—in five months: anthropic.com/engineering/re…
2
1
215
After using Sol Ultrafast, i just cant go back to anything else... even Opus feels like such a drag..
1
2
80
Sol 'Ultrafast' mode is a step change in using AI! The previous 'fast' mode was not worth it. More expensive and didn't fundamentally change how you used AI. OpenAI finally struck the right balance of price/perf with Sol Ultrafast.
72
at this point, there’s a new Decision Model launched every hour
2
69
seems @OpenAI Dots is being blocked heavily by third parties. Muse is able to do this easily. I feel over time this will extend to Muse too. UPS has already started blocking Muse.
1
1
143
enabling HDR photos on WhatsApp would have greater impact than fancy VR glasses with no concrete use case...
1
63
Really liking the dictation feature on the Muse macOS app. You can use it on the entire macOS and it's ultra fast and accurate. Definitely replaces WisprFlow for me.
Muse for Mac is out today! It works across apps, files, calendar, notes, and messages on your computer. You control what it can access. The team is shipping fast. Download at ai.meta.com/muse/download
3
1
205
Muse has replaced chatgpt for me. waay faster and more agentic for everyday use cases
62
apparently solving Navier Stokes only gets you third spot on HN.. also top comment has nothing to do with solution but around its politics..
75
if you use Fable now, are you part of the underclass?
1
43
AI needing Pretrain + RL is the perfect demonstration of "books will only get you so far, eventually you have to go out in the field to really learn"..
1
58
did openai just remove Ultra mode from Codex?? I am on Pro subscription btw.. @thsottiaux
1
1
1
130
ok found it. apparently you have to explicitly enable it now.. not sure how one is supposed to find/enable ultra mode otherwise.. also Max mode is hidden by default now..
20