mrwilson! AI for the trades β plain, practical, over the fence.
YesterYears {__} Services
Shoebox archival processing, AI keepsakes & experimentation.
{RIP}
Yesterday was a comically bad day for AI. Sometimes you just gotta take the L, another L and even the M too:
Anti-AI Anthropic
__Claude's legal guardian Anthropic had the chance to speak on its behalf with something human for once and with a msg even the anti-AI crowd could agree with: "maybe we should chill and give up on all this AI stuff" but they somehow managed to one-shot this opportunity into an overly complex, deeply philosophical dissertation paper, complete with a Mythos-designed interactive website and 20+ different outcomes to ponder more on. No mention of whether they'll want takebacksies on their potential record-breaking IPO later this year.
Monterey Park (closed)
__made it official that they will never, ever be confused with Menlo Park ever again by being the first city to ban datacenters permenantly* They celebrated their short-sightedness with a metaphorical "closed for business" sign that now "hangs" over the city to keep all the big bad infrastructure (and revenue and jobs and wifi and electricity) away. And, well, no one can argue with that.
Microsoft's Missing Form
__the company is going through survey hell right now trying to figure out how much it will cost to convince (bribe) their employees so they can fudge a few numbers in their annual employee sentiment score to look better for their boss. It's gotten so bad that they now have leagues of freshly minted 7-figure employees chasing down leadership executives with pitchforks until they add back a single magic question that got them a fat raise last year but it's mysteriously missing now. I'm not even making this up, this is literally what every future ex-employee of the company was focused on this week. However, Microsoft will spare them the embarrassment for their extreme pettiness in a simple email explaining "it's not you, it's because of AI"
L
Sam Altman is "embarrassed" that his employees are getting completely dominated in the Token Leaderboard Games, which has set off an epic pissing (away of tokens) match. There is now a 30-day challenge to find who in the world can beat the unofficial record of 603 billion (with a B) tokens spent in a month held by OpenClaw founder Peter Steinberger.
L
Meanwhile, the rest of us are trying to figure out the actual math on how much a measly 1 million in context-limited tokens on a Legacy Opus 4 model will actually cost once all the subsidies are gone and the business has spent every last dime on unfathomably ludicrous amounts of tokens for god knows what and needs our money fr fr so they can top back up again.
M
Luckily, Scam Altman still had the time to preach to all us bottom-feeders with one of his signature eerie, emotionless, lowercase-capped moments of faux reflection on X: "man the early days of the internet were so special"
Yes, yes. They very much were, Sam.
This week has been so embarrassing for the AI space, I had to write this list just to remind myself of the better days.
What is most painfully clear to me 1) the breakneck pace broke many things, including everyone's spirit 2) the messaging and leadership aren't just bad, it is an obvious root problem
We don't know yet how to fix everything. But we saw a sliver of hope in whatever Anthropic was trying to tell us yesterday. The tone revealed they are just as confused and burnt out as we are.
They were, in many ways, asking us for help.
The reason they didn't just say that is the same reason they need a $500k writer. We have to remind ourselves that this company is quite literally 80% an AI. The whole company is an AI.
Turns out, it's a really bad idea to have your engineering team run both marketing and finance. Let's start there:
- it's a brand new type of computer organization with weak people skills
- it asked us for help because it realized AI will not work without humans
That addresses a good ~95% junk of what people have been most angry about. The whirlwind hype phase was far more intense than expected. Faster and stronger, not harder and better.
It virtually started to get violent. Everyone cracked out, burnt out and at each other's throats. But we got to where we are. As quickly as we moved AI forward, we now have to go do it backwards with the "big" fix: human refractor.
So, while we assess the damage and pick up pieces from the storm, we can take a thoughtful and collaborative approach:
1) introduce a calmer period of AI (less people attacking each other)
2) allow the industry to soften its anti-ai stance (less peer pressure/shaming) everyone can learn on their own terms, at pace
3) those that can fix the broken more efficiently should offer these gains, tools or tech stacks to assist adjacent orgs/teams in getting up to speed
4) less moats, the commoditization of non-proprietary software build is very much real and virtually "free" in many operational use cases
5) genuine help and initiatives for SMBs and Main Street, the simplest AI solution for a trade or local mom-and-pop shop is many magnitudes more impactful to their ROI and quality of life
This shouldn't take a Herculean effort if we make compromises for each other.
@Polymarket Honestly, good advice. The team has been so cracked they need to press pause at the last level to enjoy the game longer. Time to get outside try something new.
@SatyaNaaksh@atmoio we are getting very predictable results of what it would look like if you have the engineers or 80% almost all computers in charge of both your marketing and finance teams π¬
All I'm asking is that they at least try a bit harder if it's what they are doing. Maybe make it look like Pratt is winning in a few of the next batches. They gotta be extra careful not to fat-finger a 24,000 - 0 now.
There is no evidence they did, as has been beaten to death here, but 4 weeks is a lot of time to potentially make an error that causes another meltdown. Why take the risk?
Just finish up the damn counting like every other state is able to do with much fewer resources. I do not understand why they leave any room for interpretation or doubt on this. It should be a non-issue.
@DilaniKahawala@alexalbert__ creating those hours-long workshop videos that no one ever watches but get shared by seemingly every single account on X, every day for 3 months at a time
No and it is a bipartisan mental issue that can affect anyone. We should acknowledge this and change the process for that reason alone. Just doing it this way why people suffer and lose trust isnβt working anymore. Our constituents should be onboard with changing it to alleviate the problem not heighten it. In person voting to get these people out offline and out of the house should be reason enough. We canβt fix everything or solve every need but when we see a problem and it can be solved by a different approach why not try it.
What is wrong is that it takes 4 weeks for some reason and without the final verdict you 4 weeks of people looking for things to do and come up with ideas of their own and then they start coming up with ideas and then they keep asking Grok to find something that doesn't exist yet and then someone believes something they saw but can't find it again and then another person jumps in to propose new ideas and on and on. We'll be in this thread for another 3.5 weeks. That is what is proven wrong about the method.
@bram Anthropic was right, we gotta slow down the AI. We need to go back to an old school to-do list for top priorities. $50 million in AI strategy for indigenous groups with no computers shouldn't even be thought of yet let alone earmarked and paid. This is crazy
We aren't getting the data for at least another 3 weeks, and by that time, it's been long enough to open the door for the suspicions of fraud. No one is gonna win here. If nothing else, it illustrates why everything was so much easier when you had to walk into the damn voting booth. Nothing to date has proven to be better any more simple, secure and trustworthy than that.
@Castantine@Geramac7@Timodc@pmarca Lollll he's got a point that sound would drive me nuts. I guess no matter how different things are today, nothing will ever solve the age-old problem of having an insufferably annoying neighbor.
This whole Cash App wand situation is the first time in my life I've felt "old old" as a millennial, and that my days might actually be numbered before having to pass the torch (err, wand) to Gen Z.
Cash App, of course, is home to a subculture of users you won't see mentioned by the brand, but are (at least partly) the main target audience the product (especially in its wand form factor) was designed for.
Hint: they aren't your Gen Z kids, but they may still call you daddy.
In fact, it's already been co-opted on social as a natural part of the campaign. You have to admit, it's a brilliant product in that sense. Cash App has a lot of experience in skirting around the topic and does a masterful job of it.
But the "old old" panic in me set in. This seemingly opens the door to willing interpretation, confusion and stereotypes, or worse. Where do we draw the line on a tech product that's being marketed to teenagers but used by sex workers?
If you had to make a comparison with AI (Grok undressing people for a few days for example), which is more intentional or poses the greater risk of long-term harm?
Whether it is Claude or the Cash App wand, we are creating such a vast web of moral/ethical complexities around the next generation's tech + culture with seemingly no way to unravel them.
It's not Gen Z's fault, it's the millennials'. It's a point now where we can't pin it on the older generations either.
The newfound old man in me says that it's time we put less focus on ourselves and come closer together for the future generations. We got realll sloppy. It's a complete mess. Now, it's our duty to figure out how to clean it up for them.
We can't make it spotless, but let's at least start putting together a laundry list.