Wonderfully cheeky but also helpfully informative writeup. Also appreciate the hat-tip to all the (as yet) unsolved security issues. Clearly MCP is onto something important, although undoubtedly the standard (or some replacement standard) will mature a fair bit before we're done with it. The flip side to that is, MCPs are probably as 'easier' to experiment with now than they are ever going to be.
They literally have no idea what they doing from one moment to the next. Obviously they are making up facts and strategies on the fly maybe minutes before press conferences. Just look all the math errors, irrational formulas and tariffing penguins. The malevolent incompetence is staggering.
These would be inherently temporary problems though right? If it became eventually clear that alternate methods were the way forward, NVDIA would be highly motivated to do the optimization work wouldn't they? Any new step functions that can forestall the asymptotic plateauing of AI progress are things they desperately need.
This is already happening. Canada is starting to accept Americans on refugee status. I have multiple US friends or acquaintances applying for Canadian Visas or Citizenship as either an active plan or as a contingency. A cousin of mine who is a Can/US immigration lawyer is absolutely swamped right now.
You know what every other online money transfer mechanism has? An ability to reverse transactions in the case or error or fraud. Because those things happen all of the time.
At the risk of taking this thread further OT, I am a product manager / consulting CPO but a whole heartedly agree with you. I like the way you phrase it. A tool to evaluate progress and provide realistic completion dates. That is how I always push companies to use agile pointing and estimating. Use it to identify unrealistic roadmap expectations as early as possible, force hard conversations on priorities and staff allocation... before committing on plans to customers or stakeholders. Or when taking a risk on a delivery date, be doing that knowingly from the beginning, with contingency plans and management owning that risk and consequences otherwise.
Rooting for Intel to survive. But shipping units for revenue on 18a is still a ways out, and if they were coming even close to signing enough foundry customers, Pat would probably still be CEO.
I went back to Rogue as an experienced Nethack player. Managed to finally win it after a good number of attempts. You have to get lucky with some early gear finds, and then spend absolutely as few turns as possible on the lower levels (otherwise, you won't survive long down there) to grab that amulet and get out. Rogue, Nethack and Nethack variants like evilhackk are all great games.
They'll have plenty of data to support the primary motivation: that enforcing arbitrary RTO policies will absolutely aid in generating staff turnover and voluntary attrition without having to payout severance costs. The policy gives them less direct control over who they lose, but I'm sure the data also points to any critical replacement employees being willing to work for less on average. That's the data they are looking at.
Indeed. Imagine Ajit Pai but now 3 steps further to the right, and emboldened to punish media and communication companies who don't do enough to promote "conservative speech". It's all in Project 2025 and beyond Orwellian. They are planning to take pages right out of Putin and Xe's media control playbook. We've already seen the likes of Besos pre-emotively complying before Trump was even elected.
A lot of these are not scams! Over the last few years I've made in aggregate well into the 6 figures of side hustle advisory calls and engagements from these networks. Legit ones I've worked with include GLG, Guidepoint, Alphasights, Tegus and i'm probably forgetting a few more good ones.
Another interesting dynamic is that if you set your rate high, you might actually see more demand. Folks who use these services are not cost sensitive. They may often pick the most expensive expert in the hope of learning as much as they can in the fewest number of calls.
Just be smart though. Don't violate any of your current employment or confidentiality agreements when taking any of these engagements.
There was a story from days of the first dotcom boom when Larry Ellison got a call from his banker complaining that he was overdrawn on his personal account by a billion dollars.
AMD also wins on perf/Watt which is pretty notable for anyone who still believed that X86 could never challenge ARM/Risc in efficiency. These days, lot of data centers are also more limited by available Watts (and associated cooling) which bodes well for Turin.
Kanban as mentioned by this commenter and others are great processes that do maximize productivity, flexibility, and often developer happiness. I love Kanban for the initial cycles of a project and as long as you can keep it going. The weakness of Kanban tends to be in predictability of future delivery, which becomes a problem when external commitments or obligations pile up, or as orgs grow and lot of cross-team dependencies emerge. So you start to lay on more structured agile process so you can better forecast future delivery and protect your team from tendencies to miss-set expectations on scope or deadlines, which can have big impacts on customer relationships or in high level company strategy decisions. So you might consciously start to trade-off in your process design, some efficiency and bureaucracy costs at the team level in favor of making better choices and efficiency at the org level.
This. This is the real social problem we should be fighting. SW should not impinge on career or social status.
As a hiring manager, if anything I'd want to consider sex performers as a green flag in a job history. Speaks to resourcefulness, social skills, courage and self confidence.
Yes! Agonizing that 10 years in and the battery still can't get you through more than one day. My pebble from 2013 was a much more practical and useable device than any Apple watch I've owned over the last decade.
Bit embarrassing for intel that they had to swallow their pride to get this one made by TSMC. But, by that same token, that alone is good reason to believe it should be highly competitive with everyone else (AMD, Qualcom, Apple) building cpus at TSMC. Particularly for it's specific niche (professional thin and light notebooks). Lets just hope Windows, and the usual OEM bloatware, don't entirely mess up it's potential for excellent battery life.
The real big lie is the idea that the recommendation algos were ever actually for the user.
The premise that many folks miss here is the idea that Spotify is, at best, thinly interested in recommending music that is good for YOUR interests. Spotify is the music business, and specifically the pop music business, has long discovered that's it's much more economically expedient to force feed musical taste onto the public than it is to chase the whims of organic hit-making. Payola is as old as recorded music. Spotify recommends what Spotify wants it's users to listen to. They have all kinds of side deals and marketing deals with labels, they have cheaper costs/royalties on some tracks than others. Popular tracks cached in their CDNs are probably cheaper to recommend than long tail ones etc. They have strategic priorities like gaining on apple for podcasts, and therefore injecting allsorts of podcast recos in the UI wether you asked for that or not.
You are using ChatGPT correctly, it is great at providing reasonably accurate background explanations on terms and concepts in complex fields. You are using it like someone would use wikipedia or a reference book, but saving yourself many trips to the library or having to search through a lot of different textbooks. Of course it's best at this for concepts that you would expect to be well documented in it's training set.
Coincidentally, while trying to get my daughter to sleep last night, I watched her solve a lengthy sudoko on my phone, while my own brain couldn't help to wander to imagining all the possible solver algorithms that might be out there. Or how I'd ever do in a technical interview if anyone asked me design one on the spot.
I got at least as far as thinking through something like your list of algorithms here. But I could help but imaging that there must also be even-more efficient or interestingly exotic solutions out there.
Like something amusing as a rainbow-table type approach where you calculate all the possible soduko boards in advance, then (somehow?) convert any given puzzle into just an index lookup of a matching solution. So like (perhaps a lot of) brute force up front, but O(1) in execution?
People are giving them grief, but you have to realize that the cost of $10 gift cards for all billion or more affected people on the planet, would quickly add up.
I think it's telling that they are delaying the microcode patch until after all the reviewers publish their Zen5 reviews and the comparisons of those chips against current Raptorlake performance.