This has saved Uber a lot of money on compute (I'm one of the devs). If your compute fleet is large and has memory to spare (stateless), performing dynamic GOGC tuning to tradeoff higher memory utilization for fewer GC events will save quite a lot of compute.
ChatGPT’s Advanced Voice Mode is one of my favorite. We travelled to Paris and Rome last year and just having a “witty, and informative tour guide with knowledge of scientific and cultural history” was very engaging, fun, and useful throughout our trips.
It’s also very useful when reading a book, especially historical ones where auditory medium works well. iPad with stage manager running Kindle and ChatGPT is my usual setup.
> Incredibly, the F91W survives its journey to an official 4,950 meters—an astonishing 16,240 feet—and back.
Findings.
> 4,950 meters under the surface, the pressure is approximately 7,227 pounds per square inch, which is well over three tons pressing on the watch. For context, that’s a Dodge Ram 1500 or a young adult hippopotamus parked on every inch of your F91W. As Americans, we’ll do anything to avoid the metric system, but using scientific terminology, we’re talking about a shitload of pressure.
As a Staff Eng, I strongly related to this. At big tech in Silicon Valley, the Tech Lead archetype does tend to do all the Project Management and, depending on whether your team can afford a Product Manager (eg. Infra team), some or all of the Product Management.
IMO, it's a little unfortunate from a productivity perspective that you have a high performing Engineer also running daily/weekly syncs, doing stakeholder management, and doing upwards management for resourcing. From a personal learning perspective, it's great though. You, as Tech Lead, learn and hone a broader set of skills and not only the Architect or Senior Engineer skillset.
I've been using Perplexity for the past 3 months on a regular basis and it has replaced DDG for most use cases. It's really nice to have answers directly.
Product comparison/updates/features:
> What's the difference in speedo latex cap vs elastomeric?
> What's new in iOS 18 Beta 4?
Documentation (esp for non-mainstream tools):
> How do i find the sioyek db files? How do I move it to my local iCloud folder instead?
I'd assume SearchGPT's results might be better given the partnerships with publishers and creators vs Perplexity searching the internet. More importantly, Perplexity already did the work of finding Product-Market Fit for OpenAI.
Perplexity is great, but they might have their rug pulled underneath them.
They use google and other search providers to run the query over the results and may be they still can find a good provider. However it's either Google or Bing and they both have their own competing products.
However, openAI might not execute this better and then Perplexity might have a chance... (I hope so).
My beef with perplexity is that there is no way to increase the number of uses of large models.
People keep telling me that I can use the smaller models, but I really can't. I'm using this for work and those things are toys which just game bench marks.
I'd love to give them an api key from openai or anthropic and get uses to my hearts content like phind does.
I see levels.fyi has a good list of "top paying companies"[1]. Is that helpful?
Side note: The difference at Staff Eng is slightly more wide. Eg. IBM Band 9 pays TC ~$250K while Meta E6 is ~$640K.
Side note 2: From what I understand, Big Tech often brings in Staff Eng from smaller companies to a Senior Eng role. Reverse is also true, Sr Eng at Big Tech are often able to join Staff roles at smaller companies.
This has saved Uber a lot of money on compute (I'm one of the devs). If your compute fleet is large and has memory to spare (stateless), performing dynamic GOGC tuning to tradeoff higher memory utilization for fewer GC events will save quite a lot of compute.