The Decline and Rise of IBM (1996)
sloanreview.mit.edu
sloanreview.mit.edu
It went off constantly with a certain "type" of "startup executive", and I noticed they were all former IBM or Intel. I'm not a negative person but found them all to be visibly political, all sounded like that didn't actually do anything really real, and talked a big game that the company outputs didn't seem to match.
I guess they are costing the quantum startups (aka VCs) a lot of money for very little actual work. This is just my observation and just my experience but I'm keeping tabs because it will be interesting to see how many quantum startups fail over the next few years, and which were run by IBM execs. I have my hypotheses ready.
Its kind of interesting. It's such a huge organisation that there's room for multiple different streams, trends and directions all the time. There's also a massive disconnect between the people producing tech products, the 'research labs' and the services arms, the latter of which seem to have the focus of upper management more than any others.
The 'labs' are there AFAICT mostly to make the company look good and try to achieve headline-grabbing things like Watson. Very little of this feeds through to the trenches of the tech production lines, which tend to vary from the obvious corporate/enterprise software type setups (monstrously huge, cumbersome software which does things useful to enterprises but takes a long-ass time to adapt, gets its lunch eaten by startups, which then get bought, replace the incumbent, start to grow ... rinse and repeat), to large departments which resemble nothing so much as academia, complete with dingy corridors that haven't been redecorated since the 80s, which have miniscule output expectations and release new products with an absolutely glacial cadence.
Meanwhile the real corporate focus is on selling services and consultancy.
I feel somewhat fondly towards IBM, and I'm glad to have worked for one of the biggest names in our industry. They always treated me well (though make no mistake, they'll lay off thousands in a heartbeat if they think they can save a dollar). At the same time the organisation did feel a bit pathological, particularly last time I dipped my toes in, it felt like a giant in managed decline. Still vast revenues and enough money to cast about looking for lifelines, but ultimately directionless. Perhaps things have changed since 2015... perhaps the redhat acquisition has altered the focus. I'm guessing not though.
Right?
That’s why everybody is raving over how awesome Bard is!
Sure, there's things that can be improved, but years later they are still the most ubiquitous binary message format.
I feel there's no one trying to take that crown really.
Even the integer encoding and tooling is very similar, just not quite identical.
Golang seems bigger outside of Google than inside, too.
(Like the nick, BTW. Wish i'd never sold off my TX81z)
A nontrivial fraction of all living people have heard this song at least once, and it’s 100% TX81Z at the heart: https://youtu.be/9EcjWd-O4jI?si=XgkC5avENfdTmlQF
(Also, yeah, to other replies, I now realize protobuffs are not to be joked about, I repent. Forgive me.)
A long time ago I wrote a little FM synth in C++ and I went through some painstaking effort to copy Lately Bass and actually got pretty close. It's all about the envelopes. That code is lost to time though.
I think I still have a Yamaha TQ5 downstairs, which is basically a 4 op in the same vein as the TX81Z but in a form factor that looks like a 1980s alarm clock or something. Very ... shoulder pads and avocado bathtub era: https://yamahablackboxes.com/collection/yamaha-tq5-synthesiz...
Also, if you can get your hands on Korg Opsix it is a lot of fun for FM. Does about as good a job at making FM closer to a subtractive synth in terms of ease of use. Of course, FM will always be way harder to deal with than subtractive, but it’s still the best attempt I’ve seen.
JSON is wildly imperfect, and lacks types and schema enforcement... but if I have a stream or file of it I can read it with my eyes. Protobufs are opaque blobs without the schema and a pile of code. And they use generated code which is on the whole rather annoying and of varying quality. The binary encoding of protobufs doesn't really give them much of an efficiency edge, either; in this day and age there are some very well tuned JSON etc parsers. And if you're looking for efficiency, there are other formats better than protobufs (or thrift) for that purpose.
At Google I saw it rammed into places where it really didn't belong. Embedded systems, even, causing months of delays while people struggled to get them to work acceptably on a microcontroller class machine. Dogma.
But at least there we had better tools inside Google to work with them than I've seen outside.
TLDR they're probably more good than harm, but I don't think they're a universal good and not the universal answer for how to send messages "between things"
I work in Rust and use binpack, flatbuffers, and json, or hand rolled binary layouts (often using the binary_layout crate), depending on the use case. I don't think there's a universal solution.
They landed on their feet.
So I ask again, what does IBM actually do?
It’s hard to describe the impact of that time - IBM centers were eliminated or eliminated thousands of jobs. Places like Kingston, NY and Binghamton, NY just died as the mainframe business pulled back dramatically.
Everyone laughs about stupid IBM, but frankly they are the only meaningful survivor of the early days of computing and find a way to stay around.
Power10: https://en.wikipedia.org/wiki/Power10?wprov=sfla1
No page for Z16 so here's the Telum processor: https://en.wikipedia.org/wiki/IBM_Telum_%28microprocessor%29...
The wicked thing I got to work on is the Explorer memory chip. Crazy fast throughput and lane width, truly. 38Gbps for DDR4 RAM
The rest of the company could all disappear and I don't think anybody would mind all that much. I've never heard of anybody passionate about working with or on anything besides the core systems. It all looks like consultancy and sales bullshit from the outside.
Companies that depend on big government contracts, and "consulting" (quotes are deliberate here) don't ever rejoin the small business and consumer sectors, ever. They're too big, bureaucratic, and slow.
If you're trying to hold up government and Fortune 100 companies as somehow more virtuous than small-to-medium businesses, you can't be serious. At least they're not too big to fail.
here in California over decades, income and empires were often consumer-based products with consumer distribution chains.. huge money.. the halls of upstate New York and the skyscrapers of Manhatten were viewed as distant and out of touch. IBM and maybe others were household names, but the source of revenue and the way to participate was completely alien.
So I think you have hit on a nerve, yes.. there is some disrespect
The old saying that “revenue is for vanity and profit is for sanity” never rang so true.
Two examples I was close to:
Interleaf. Interleaf was the first good commercial document processor.[1] Released in 1981. Ran on Sun workstations, quite well. Years ahead of Microsoft Word. The business model was that they would sell you four workstations, a server, and a laser printer for about $60,000. In my days at the aerospace company, we managed, with great difficulty, to get Interleaf to sell us the software alone, since we already had Sun workstations, a laser printer, and a phototypesetter. (Aerospace generates a lot of documents with diagrams.)
As a software-only company, they could not have obtained funding, or, probably, generated enough revenue to operate.
AutoCAD. The story of AutoCAD is well documented in The Autodesk File.[2] Autodesk was founded in 1982. A software-only company was not fundable back then. Autodesk started off with about $62,000 put in by the founders. (Market cap today, $47 billion.) The term sheets for some VC deals that didn't happen are in there.[3] "The overall flavour of the deal seemed to us totally inappropriate for a company which was, at the time of these negotiations, generating sales equal to the size of the deal every month and generating after-tax profits close to the size of the deal every quarter." Autodesk couldn't get funded on reasonable terms even after becoming wildly profitable.
It was an interesting time. There was so much that obviously needed doing and hadn't been done well yet.
[1] https://en.wikipedia.org/wiki/Interleaf
Ashton-Tate was large. Software Publishing (PFS: File) was large. Word Perfect was large.
People didn't think like that back then. They wanted profits. The idea that a company's value can be measured by their burn rate is 21st century.
Sevin Rosen Funds, out of Dallas, is forgotten now, but they funded Lotus and some very early software startups.
[1] https://www.bloomberg.com/news/articles/2023-04-19/ibm-earni...