1. If it's not written down, it does not exist. This needs to be a ruthlessly enforced rule. I know we all love those Mel the Programmer stories, but that's a story about fragility. Similarly if you are relying on SRE or dev heroics to solve field problems, you have a gun to your head and you need to solve it yesterday.
2. The TCO of knowledge is dominated by the maintenance cost, not the creation cost. I see this mistake made a lot. Anyone can create a confluence page, Slack comment, SharePoint site or a git comment. The org must make it someone's job to grow, weed, and maintain these gardens. It does not matter what system you use, but someone needs to look after it.
You get a bonus for building information "data structures" that make it easy for people to pull data out as your org changes, but this is probably something you should accept that you'll get wrong as your org grows. AI hides this problem because it makes search so easy, sometimes, but it will come and bite you hard one day, so be ready for it.
The UK power grid operators famously plan for massive demand surges at eg the end of major football matches. I cant imagine what their forecasters thought of the England-Mexico nail biter. Half the nation heading to put the kettle on, half glued to their seats. Would love to see the charts of demand now that the World Cup is done...
And inversely, steer away from (named competition).
If I was Corp A competing with Corp B, what stops me, today, from flooding reddit etc with anti-Corp B material to pollute the model, and then paying OpenAI to not just nudge users towards me but also away from Corp B?
This is an odd little device. It seems you can only call other express chat users? it's reminiscent of cellphone systems from the old old days like Nextel walkie talkies. Also I see its available in India only and given the feature restrictions perhaps it's meant as an ultra low end device ?
If HMD came out with something that was optimized to the nth degree for whatsapp calls, that might be interesting, but even then I wonder how cheap it would need to be to compete with budget androids.
I'm going to mangle the terms of art here, but the ur-problem is that labels of routing, like phone numbers and email addresses, get confused with labels of identity, and then with indicators of trustworthiness.
Everything is built to address that weakness - think DKIM, SPF, etc, plus STIR/SHAKEN, to say nothing of IP or ASN filtering, but they feel like bandaids on a very difficult problem. What you end up with are basically default-deny except for a personally curated trust set ("only accept calls from my contacts", "everything goes in spam unless I have previously corresponded with the sender"), etc.
One last robocall story. AT&T sat on their hands for years until consumer groups embarrassed the then-CEO enough to do something about it. There was a memorable interview in the Dallas Morning News where they called him on it instead of lobbing him softballs and I suspect that the embarrassment finally got through.
I sometimes feel that when the industry moved from pets to cattle, what really happened is that the cattle turned out to need an exotic farm to live on, negating the savings. You can have a few honking servers or you can hand massage exotic k8s setups on your fleet. Pick your poison, but dont delude yourself that the TCO of the latter is lower than the former.
I worked on STIR/SHAKEN for the two biggest US operators. The techies tried very hard to make it work, and, indeed, there was a brief time when it worked pretty well, but, the incentives from Corporate were and are fundamentally misaligned.
Type A attestation is, generally, solved. Carrier A attests that the number is one of theirs, and they know that the caller is one of theirs too and attached to their network.
However: this is a fraction of calls. Carriers also sell blocks of phone numbers without the corresponding access network. This is what allows you to pick, say, a Twilio number with a local area code. In these cases the best that can be hoped for is a lesser attestation.
But it gets worse, because the operator can also sell blocks of numbers to people with no direct connection to the US carriers and who need to spoof US numbers. That call from Capital One comes from the Philippines via two or three intermediate operators, none of whom can attest to much of anything. And into that gap ride the spammers.
Furthermore, in an Experian-like twist, some carriers also realized that businesses would pay to have their calls show up as "trusted" on the recipient's phone. So the standards were enhanced to deliver 'rich call data'. However, in order to be something worth paying for you also need a baseline of calls that do not have that premium look. A scam? You decide.
Finally, one other misaligned incentive. All of this needs VoIP. Not TDM (classic legacy telephony). However, the big US carriers make bank selling TDM circuits to the hundreds of small regional telcos, and refuse to sell them SIP trunks, because it's s such an easy money maker. So again, technology loses to incentives.
These incentives, to make money from phone numbers, vastly outpace what
The problem with (Enterprise, 365) is that it's critically underpowered for precisely the kind of knowledge work that the 365 power users need it for. Summarizing emails or Word docs, or creating a ppt from an email, are essentially parlor tricks. The value is in the long context that encodes your specific business practice...except that copilot flakes hard when the window gets long, and can only be fed a handful of knowledge sources.
Copilot also has no semantic knowledge of Microsoft-but-not-365 things like PowerBI dashboards, so your average cubicle jockey, who doesnt have access to the data model, can't, say ask it to pull last month's sales figures. That's a giant hole in capability.
The agent discoverability is non existent, so teams end up repeating work. It doesnt have a skills.md type of model that allows for easy, non-admin, extensibility. And most fatal of all IMHO is that it does not play nicely with other non-Microsoft things out there. This may be as much a fault of Salesforce and Atlassian etc as Microsoft's, but Copilot cant see into and manipulate objects in those products (certainly not without admin-level approval, and even then, not very much).
I feel that Microsoft are losing the game here by not allowing a genuine democratization of what the tools can do. Excel is impregnable, not just because society is built on it, but because even the lowliest office grunt is allowed to use all of it. If Copilot can't be used and wired to everything that a knowledge worker uses without requiring IT- or 365-admin approval, then what will happen is that Shadow AI will show up and workers will start bringing thwir own AI tools into work, with ugly consequences for security and data exfiltration.
My crystal ball says that Microsoft know that CP cant do it all, so they will pitch Cowork, as a sort of workaround, at which point some companies will balk at the price and park CP in a dark corner while the real work gets done in Google Workspace or Slack or whatever other environment gives them more of what they need.
Cuts two ways. Why should I pay $200 for a BigBrand dog bed if this knockoff site shows SHRDLU has the same thing for $40? We all know that BigBrand gets it from the same supplier.
The real knockoff problem I see is that you buy what you think is BigBrand and get shipped Knockoff because someone is mingling inventory.
I wanted to sympathize with OP, but I sense that the blog post was written while they were still fuming, and some of their behavior during the process (esp the first job) gave me pause. Maybe its just the timing that makes it feel aggressive.
Imagine you are a contractor on a gig. You don't wait until things break before you ask for access. You ask, before the job starts, what all the doors are and who has the the keys. So: "hey, Recruiter, in order to make this task a success next week and make you and me both look good I will need some stuff ready on day 1: what's your code repo system ? MFA?internal comms channel? bug system? internal knowledge base? and most importantly, if it's not you who can administer access to these things, who is the person I talk to, and can you arrange an intro on day 1?"
To not show this level of initiative, and sit around while your friend navigates the corporate machine for you, is the behavior of a new intern, not a senior developer. OP also doesnt say that they arranged a live end of Day 1 checkin with their task manager, which is another miss, since it allows for the latter to mop up any lingering issues ("oh yeah I forgot you need access to X, let me sort that out") and for the former to remind them that this is an odd task ("it was interesting that you asked an Elixir dev to work on a browser extension")
As far as being asked to work in TypeScript not Elixir, thats certainly odd, but could have been deliberate and not necessarily foolish. Maybe they wanted to see how intellectually adaptable OP was. Maybe they wanted to see if you would speak up and handle potential conflict. Or maybe they had a mixup between two candidates. We'll never know because OP sent a wall of text which surely would not have helped defuse the situation.
Amen. Normal office work is wildly different from what we read about on HN. If you were a CEO, determined to lay off all your people, you would want to really zero in on having your AI solve these very unsexy problems: extract data from Office and PDF. Grab data from some part of the screen of a webapp and parse it. drive a line of business app via keyboard or mouse simulation. I know there are companies out there that try, eg Appian and (here in YC) Skyvern, but its a hard problem and yet I feel this is where the true money is.
I dont see this experiment as any kind of "failure". Something was learned, and OP is better off for it. Computing and science literature would be a lot better off if people, like OP, honestly documented where things went wrong.
The text that SCOTUS put out today discusses this in great detail. You might agree or disagree with their reasoning, but they consider what "obvious" means in light of the text and the writings of the time plus references to old commentary like Blackstones. Maybe better to read that and assume in good faith that this is what the commenter is referring to?
I do wonder how Microsoft will manage the transition of the NT generation. Raymond Chen has been doing this kind of work for thirty years. He probably has, what, another ten years, max? Who are the next generation of Windows gurus that will take up the mantle?
Hanselman is good on the blog part, but not in Chen's class re Windows domain expertise. And Windows is not, despite all that is said, going away anytime soon. I think this could be a real problem.
Hopefully there is a set of 25 year old developers in the Windows team who have deep and growing skills in Winternals, and Microsoft have the good sense to encourage them in their career.
The problems isnt LLMs per se, it is the shift to trusting the output of the machine coupled with a decline in verifying that the output is reasonable. It's basically what your teachers warned you about with wikipedia in eight grade except applied to all areas of life, including medicine. Dictation is already high-stakes and LLMs do not automatically reduce that risk.
Here is an example. My provider sent me this note. I'm quoting verbatim here from my MyChart record:
"Your liver enzymes are high, I would like to order acetaminophen containing medication like Tylenol, I would like to order liver ultrasound I placed ultrasound order in the system, make an appointment for radiology, I would like you to get hepatitis panel lab work done, obtain blood work order, please schedule a well visit to get it done"
When I queried it, this is what I got back. It was a dictation error. You could almost hear the panic in the message:
"Sorry for wrong message earlier, I was dictated message- so could not realize that it was written to take Tylenol type of medicines- I DO NOT RECOMMEND ACETAMINOPHEN CONTAINING MEDICINE - LIKE TYLENOL AND ALCOHOL DUE TO ELEVATED LIVER ENZYMES."
Again the problem is not dictation, or LLMs. The problem is humans ignoring their responsibility to check the output of a machine.
Opinions expressed are my own, and not those of any employer, whether past or present.