Shaheen Gandhi

might actually have to become a united customer again.

@mitchellh transition to azure is rough…

Every. single. day. It's increasingly becoming difficult to do real work with GitHub. Git isn't the issue, since I can work offline. Its issues, PRs, CI, etc. Imagine going to work and your workstation randomly restarts a few times a day. That's what it feels like using GH.

@pitdesi oof Seattle :(

I thought "lemme see what the data says" but it was much harder than I expected because Miami has actually fallen off the chart...

@pitdesi @bscholl smisek-era united-continental was pretty bad, turning an amazing airline into a subpar one. i might give united another chance now.

IMO United is now the best US airline. But if you live in a city that is a hub, just fly whatever airline your city is a hub for! Will make your life easier.

@jasoncrawford my bad answer is “claude’s girlfriend” but @bcherny will hopefully tell us…

Um… Claude Code just created a .claire directory? For its git worktrees. Who is Claire?

@pitdesi paging @dieworkwear

I could see myself wearing this everyday. Why is there so little crossover between comfort and style?

the environment allows you to write and test your solution locally and then submit when you're ready. bonus if you have claude code: it will generate test cases for you from the problem description

i started playing around with @LeetCode. being a terminal junky, i wanted to do everything from the command line, so i found leetcode-cli. then, i got bored waiting for test submissions to run, so i created an offline environment: github.com/visigoth/leetc…

paging ⁦@PalmerLuckey⁩ instagram.com/reel/DS_xzR4Dc…

@ID_AA_Carmack @Jonathan_Blow the most recent counter example is Xbox, where ripping out a lot of Windows helped us gain technical credibility within console development. i had rose-colored glasses from that experience, and have been involved in every make-a-consumer-client-OS effort within Meta.

I had been meaning to comment on @Jonathan_Blow’s “Why can’t we even conceive of writing a new OS today” post. Coincidentally, I just got an email that opened with: ——————————————— Hey John, 2 years ago I pitched you LIBBA - a dedicated OS for smart glasses. You were sceptical, your main concern was that a custom OS rarely justifies itself: cost, shelf life, and developer burden outweigh the benefits. You were right. ——————————————— I deeply love the ideals of clear, efficient programs that do their job without baggage, and I have always been very sympathetic to efforts like Oberon, Plan 9, and even TempleOS. But building a new operating system today doesn’t make any product sense. Meta spent a lot of resources working on a fully custom XROS, over my rather strenuous objections. They had top tier engineering talent, tons of support, and they were producing high quality code and docs. It was a best case scenario from a “new OS” perspective, and, as one of the engineers put it, “If we can’t do it, who could?” I wish I could drop (so many of) my old internal posts publicly, since I don’t really have the incentive to relitigate the arguments today – they were carefully considered and prescient. They also got me reported to HR by the manager of the XROS effort for supposedly making his team members feel bad, but I expect many of them would acknowledge in hindsight that the Meta products would not be in a better place today if the new OS effort had been rammed into them. I can only really see a new general purpose OS arriving due to essentially sacrificing a highly successful product’s optimality to the goal of birthing the new OS, and I wouldn’t do that myself as a stakeholder. To make something really different, and not get drawn into the gravity well of existing solutions, you practically need an isolated monastic order of computer engineers. Which was sort of Plan 9…

@openInfer when can i replace my ollama instance with openinfer?

🚀𝐎𝐩𝐞𝐧𝐈𝐧𝐟𝐞𝐫 𝐉𝐮𝐬𝐭 𝐆𝐨𝐭 𝐅𝐚𝐬𝐭𝐞𝐫: 𝟐-𝟑𝐱 𝐏𝐞𝐫𝐟𝐨𝐫𝐦𝐚𝐧𝐜𝐞 𝐁𝐨𝐨𝐬𝐭 𝐎𝐯𝐞𝐫 𝐋𝐥𝐚𝐦𝐚.𝐜𝐩𝐩 / 𝐎𝐥𝐥𝐚𝐦𝐚 Why it matters: OpenInfer runs 2-3x faster than Llama.cpp & Ollama on 𝐃𝐢𝐬𝐭𝐢𝐥𝐥𝐞𝐝 𝐃𝐞𝐞𝐩𝐒𝐞𝐞𝐤-𝐑𝟏 (𝐐𝐰𝐞𝐧𝟐 𝟏.𝟓𝐁, 𝐐𝐰𝐞𝐧𝟐 𝟕𝐁, 𝐚𝐧𝐝 𝐋𝐥𝐚𝐦𝐚 𝟖𝐁) 🔹 Benchmarked on AMD Ryzen 9 7845HX, significantly higher performance on Distilled DeepSeek-R1 models 🔹 Optimizations: Smarter quantization, faster memory access, model-specific tuning Next up: 𝐆𝐏𝐔 𝐨𝐩𝐭𝐢𝐦𝐢𝐳𝐚𝐭𝐢𝐨𝐧𝐬 for even greater speed. #AIInference #EdgeAI #MachineLearning #Performance #OpenInfer #DeepSeek #Llama3 #AIModels #Benchmarking #EfficiencyMatters

@jmitch @pitdesi the bootloader is open, so you can just boot whatever os you want on it already.

@pitdesi No it’s totally useless. I want them to open source it so I can build a dumb phone for my kid to use. Otherwise it sits on my desk in the case 100% of the time.

Anyone who got the Rabbit R1 finding it useful for anything?

@sdamico what are the major micros that do battery management and motor control?

4) scale from EVs and drones led to purpose-built single-chip solutions for battery management and motor control, driving lower costs, better overall integration, and letting new companies benefit from abstraction when using these technologies.

FOUR key things in the past 20-30 years are causing the physical world to electrify: 1) magnets got way better, enabling high torque permanent magnet motors.

@NielsHoven i don’t think the verbiage is stating that is the reason. it is just stating that kids experience diversity at public school, which gets lost when moving kids to private. but, that’s just one attribute and for the quote to single it out clearly makes an implication

No, parents are not leaving public schools to avoid "experiencing diversity". Stop trying to turn everything into a culture war. Parents are leaving public schools because out-of-touch, anti-merit policymakers have deprioritized excellence in our schools

@pitdesi i can say i only understood the framework here due to @patio11 ‘s recent essay on interchange

Wowwwww Everyone in fintech has been saying that the Bilt credit card ($3.1B valuation) is too good to be true; How can you get cash back on rent payments? Turns out they have an unsustainable deal with Wells that is losing the bank hundreds of $millions. Wells made a number of incorrect assumptions in their underwriting. They assumed that it would be a top of wallet card for renters who carry balances. Instead, a lot of savvy consumers (myself included) use the bank for the minimum required to collect the rental points. Wells assumed that 65% of card-purchase volume would be non-rent, generating interchange-fee revenue. The reality is inverted... most of the volume is rent payments on which they don't earn much. Wells also assumed that 50-75% of the balances would be revolving, earning them interest. The reality is they mostly have cardholders like me who don't carry a balance - only 15-25% are revolving. Wells is paying 0.8% to Bilt on rent payments, even though the bank doesn't get interchange on them. Wells has a massive mortgage business and thought that they'd be able to cross-sell mortgages when the renters become homeowners but that hasn't panned out. Kudos to Bilt for negotiating a killer contract that doesn't expire until 2029... but I wonder what the future looks like after that. Wells says they won't renew it as-is. In the meantime - it's a great card... pay your rent via the card and make 5 other transactions a month to get 1% back on your rent.

I’ve been doing negotiation coaching for tech managers, execs and entrepreneurs for 13 years. Here are 5 of the most important learnings I’ve found:

@adityaag i believe jensen runs NVIDIA this way - i've heard that, generally speaking, there are no extraneous 1:1s

4/ This doesn't mean feedback isn't useful. But it should be given every 3-6 months, not weekly. This forces managers to identify patterns and provide holistic guidance, rather than doing weekly "spot checks."

1/ Weekly 1:1s with direct reports are a staple of Silicon Valley management. The idea is to check in, see how they're doing, and provide feedback. I did this for 10+ years at Facebook & Dropbox. Frankly, I hated it and found it useless. But it's what "good" managers did.

@bzotto doc brown would have

it unfortunately came up with tautological answers to those follow up questions. but, the code was correct which is more than i can say for most candidates!

i have done almost 500 interviews at facebook, and while i don’t ask this question every time, i’m well-calibrated on how candidates perform.

this morning i asked ChatGPT a programming question i ask in interviews at facebook. it did better than most candidates.

once it did, the solution it gave was pretty good. it was functionally correct. it also landed on one side of a small design decision that is worth understanding (when to return an error or not). i asked it to justify the design.