I worked at Facebook and remember what it was like to first face the total complexity of the system.
It’s really hard for me to imagine myself working at a company that deals almost exclusively with machine-generated data (as Tesla does) and being assigned to judge a social platform’s engineering qualify in one day. Cars and spaceships on a closed network are fundamentally “clean” compared to having hundreds of millions of people who produce dirty and often adversarial data.
But I guess the point of this exercise isn’t really to judge code quality, but to drive fear at Twitter.
I suspect it’s less a complete review and more an interview of the senior technical manager, to get a sense of whether they are up to par. Do they think about observability, alerting, refactoring, service architecture, etc. reasonably well? Are there issues with how they motivate their teams or when they give them responsibilities? Are timelines challenging but doable?
“Code” is most likely “architecture” and “architecture” is code for “management”.
They're different... Twitter software engineers get very high pay to do dull work, whereas Tesla engineers get actually impactful work, but very demanding working hours and low pay.
Of the engineers I've met, I'd put the Tesla ones as 'smarter'.
This is strictly re: software engineers but I don't think that's true IMO. Put another way, I think the average recruiter/co-founder/EM would be more interested in hiring an ex-Twitter engineer vs an ex-Tesla engineer.
but I have inspected plenty of organisations where strong engineering teams are drowned by management bullshit. even found places where most of the individuals, engineering and management both, are arguably competent but the overall organisational decreptitude kills any and all effective work, no matter how heroic or skilled the individual and collective effort gets.
Honestly, it's mostly because they pay more. They're more selective and thus have better engineering talent meaning Twitter engineers on the average are going to have a better engineering skills than Tesla software engineers
Tesla engineers likely are drinking the Kool-aid or were not talented enough to get offers at better companies.
Lastly, it's not "just tweets" at their scale. That's like saying Netflix is just storing some video files in CDNs.
I think it’s less about strictly better talent and more about trust and familiarity. Musk trusts Tesla software engineers to tell him what is good, what is better than Tesla, and what is concerning.
I’m not a software engineer (data scientist) and I’ve only managed up to 15 people, but if a friend who worked in a factory asks me to review a CTO of a team of 50, I feel confident I can say “Asking people to write their own unit tests isn’t abuse of power” and “Asking people to be behind their desk at 9 am or risk pay cut would not fly anywhere else” — things that he might not have the context to know.
Sure, there might be a new paradigm that I’m not familiar with where devs don’t write unit tests, but I’m fairly confident if I hear about that, I can ask questions and see why it’s more efficient.
A really good (reverse) example of that is James Douma. Several Tesla commentators rely on him to comment on Tesla’s AI announcements. He doesn’t work at Tesla and confesses that he might not be able to lead the AI team there, but he’s personable, clear, and familiar enough with ML to offer some judgment on why things are done in a certain way. He’s usually very positive but he could easily be asked which teams don’t seem to be executing as well, and what questions to prod them.
Auditing what you just bought is about as smart as taking the used car you just bought to the shop to get everything checked over and tightened and lubricated.
... got the money, then found out the previous owner was a chainsmoker and a cadaver is decomposing in the trunk, then for some mysterious reason didn't want to buy it.
He did want to buy the car, signed a contract to buy the car that decided that he didn't want to buy it because it had a head gasket leak. Even though that was the whole reason he was buying the car, and he doesn't have any evidence for said leak anyway.
That’s precisely how acquisitions at my current and past company worked. Everyone signs NDAs, there’s a fee paid by the purchaser for the time should they decline the deal, and the audit commences.
I’ve been asked to do the evaluation a few times and it’s pretty straightforward. Even if you think the code is of poor quality, it may still make sense to complete the purchase because of the business case.
Not necessarily. It depends on the company acquiring. I've seen M&As focus solely on the relationships (ex, customers), so the tech DD was just checkboxes without deep dives.
Not necessarily… what? This is my actual professional experience. I’m sure people just buy things without any due diligence somewhere but I have never seen it.
In one case my judgement was that the customer list was all that there was of value. They opted to not buy that company.
Getting/giving access to the actual code/backend (especially as a competitor!) is a heavily negotiated diligence request.
The seller can simply refuse without even giving a reason.
ITT:
People are throwing around the phrase "audit" liberally. Audit will have a defined (and likely limited) scope and is typically more about compliance, e.g., Do you have the correct number of Microsoft Office licenses per accounting records?
Inspecting the source code is a condition of all deals I've been involved with. Most sensible people want to see what they buying, unless they are under pressure to close fast due to competition. It's a lot like buying a house.
I once read about an acquisition attempt by Google which eventually failed because Google found codebase less than impressive. So I guess this is normal to review codebase before a buyout.
I utterly fail to understand how this would yield anything other than.
Ooh they use tabs here rather than spaces! And that experimental code over there contains an unused variable! This A/B testing pipeline has minimal test cases! They ship in a day without a Q/A pipeline.
All of the above would happen if any company were acquired by any other. The code style will be different, and the dev practices will be different. Companies optimized to ship quickly will appear low quality to those who optimize for quality, optimizing for quality will appear as a slow moving dinosaur to those who move fast.
Why? One of the production goals of software is to minimize inventory time. Software that is written and not shipped incurs capex without revenue, unlike tech debt - I can put an explicit number on this cost. Similarly, software that goes unshipped also does not get customer feedback. It may be worthless.
There’s a balance here - but it’s not at all obvious what the balance should be in every business.
You're right, every business is different, and that goes for software businesses, too. I'm thinking that Twitter's central position re the dissemination of information, and its international scope make it a fairly high security operation, in this day and age, whether it wants to be or not.
Yes, software is typically more complex than a laymen would expect once all of the edge cases and layers of legacy code or “guano” are implemented. However, I doubt that these outside developers are expected to understand everything in a single day.
There are legitimate questions regarding how certain sections of the Twitter code operate as it relates to censorship, which is a primary reason Elon purchased Twitter. Given the activist mentality (and possibly hostile to new ownership) of a certain number of developers, it seems perhaps prudent to analyze the code and evaluate the existing developers immediately upon purchase of the company.
In a typical acquisition this would have taken place under NDA during due diligence, before closing the deal. But Elon waived it.
Now he’s apparently trying to find this shadowy “fifth column” of engineering over the weekend so they could be fired before RSUs vest on Tuesday. It sounds completely ridiculous but isn’t entirely implausible since, after all, we’re talking about the same person who committed $44 billion without due diligence.
The “censorship” aspect is pretty interesting because nobody seems to care about retaining existing users. HN has much stricter moderation than Twitter. Imagine someone buys this site tomorrow, fires dang and lets conspiracy theories run wild. Would you stay? I wouldn’t. And a social site is only as valuable as its users. If the new owner of HN promised he’s going to make this site into a “super-app”, it wouldn’t make any difference — I wouldn’t come back, just as I’m not going back to MySpace or Twitter.
Nobody in Elon-space seems to be bothered that there’s no undo on alienating a social network’s user base. I guess they expect to get a billion new users somehow (who also want to pay for the service since that’s floated as a business model for Twitter). But to me, this whole deal feels like Tumblr joining Yahoo.
If Musk would have announced the details of the layoffs within 24 hours of closing the twitter deal, commentators would have said, “So soon after taking over it is be impossible to make an informed decision about how to restructure and who to keep and who to layoff.
If he uses the best means of making an informed decision (use capable engineers he trusts and that are outsiders to twitter) and takes some time to properly consult, commentators say, “It is complete chaos and an information vacuum. He is just driving fear.”
It is right to closely watch people in power. But it doesn’t absolve each of us of the responsibility to hold us to similar standards when commenting. Most commentators come across as being in a hysterical, and not in a judicious frame of mind.
What you outlined were not the only viablw options: Musk had months to do a proper transition after signing the agreement. He chose not to do that, for reasons best known to him.
It should be expected that when you force yourself into a bad situation, and can only make bad decisions, people will observe that you made a bad decision when you inevitably do.
It sounds like your not a fan of Elon and nothing he’ll do can change your opinion. Why not hold back the strong judgements and conclusions for at least a few months? Maybe he will surprise you?
See, this is where a car and a social network are fundamentally different.
I own a Tesla. It’s a good car (although I’m unhappy about how much I paid for the FSD package which was nothing of the sort — borderline advertising fraud). Next year I’m probably buying another car, and if Tesla is still the best for the money, I’ll get one regardless of my personal feelings.
But with Twitter it’s not an objective decision. The app doesn’t deliver quantifiable day-to-day value like a car does. Now that I’ve removed Twitter from my life, it’s very unlikely that I’d ever go back.
MySpace did a lot of product changes in 2008 to bring back the users who had defected to FB. Nothing moved the needle. That’s just how it is with these products. I’m surprised Elon pretends to be oblivious to this dynamic.
My 10-year experience with Twitter was one of frustration. I had almost 1,000 followers, I tweeted regularly, tried to be nice and occasionally witty, replied and retweeted mostly useful stuff. Yet I never got any engagement. A few likes on a tweet was the upper limit of interest. Every other social media platform is much better at rewarding regular users like me. But Musk doesn’t seem to want to address this; instead he wants to bring back Trump. Why should I stay?
Hard for me to say. What type of reward are you expecting to get out of Twitter or social media in general? Maybe you are expecting too much or the wrong types of rewards?
Why not avoid the algorithmic feed from Twitter and only look at what your chosen followed users are posting? Relying on a third party to provide both breadth of information and inoffensive posts is asking too much.
Just as in real life, not every person you cross paths with is worth your time. Wading into public spaces like Twitter requires some work on the users part to find gems of useful information … relying on the platform to do it for you is likely to never work perfectly.
> I notice that Musk is spending his time tweeting that it is possible Paul Pelosi was attacked by his gay lover, so it's not looking good.
I don't have twitter, and this narrative seems like the work of 4chan (4 the lulz), but it's true he really was hinting at this [0] being the case.
I just hope this cesspool dies already, it seems liek an echo chmaber for the most feeble minded. I was catching up on the all in podcasts (missed 3 weeks worth) and the amount of servile rhetoric for Elon from those 4 is absurdly out of touch with reality given his actions--it's clear he was forced to overpay for something that has little utility in the grand-scheme of things--henc why Dorsey walked away. If this is to be successful, perhaps this tech apocalypse will start the demise of these platform's and will have less relevance from this point onward.
But, I feel that he is doing what he is best at: Marketing. His tactics to bring in ad revenue are the same that cable news has with polemic based click-bait but that sad reality is that he has a legion of simps to dog-whistle to keep it afloat.
Other publishers, substack, etc. Why would anyone actually need to be on twitter? so they can break up their insight into 280 character bites and post them in 30 post-long threads that only other twitter users will read and eventually be re-digested into something most people will consume? Vine users didn't go anywhere for the longest time, most of them outgrew it by the time TikTok came around. Twitter was a fun experiment but it seems like it's been on its way out. Seems like just about anything that Elon is going to want to do it will result in deflating whatever is left.
> the same person who committed $44 billion without due diligence.
Honestly, I'm not sure what due diligence you think he should have, or could have, negotiated. The code obviously works. The userbase is there. They are public, so a lot of information is known.
> Given the activist mentality (and possibly hostile to new ownership) of a certain number of developers, it seems perhaps prudent to analyze the code and evaluate the existing developers immediately upon purchase of the company.
Checking the code doesn't make sense, unless you think Twitter is some rinky-dink outfit[1]. If they want to inspect foe activism, they should check the configurations - not code. The engineers only add the knobs and dashboards[1], an entirely different team takes care of operations.
1. Based on what I've gleaned from similar-sized tech companies, or even those 2 orders of magnitude smaller.
My guess is this would be a very high level look at their architecture and maybe a closer look as some specific portions. Like you said these systems are so complex that you wouldn't be able to figure out what is even happening in a day.
I don't agree that it's to drive feat at Twitter, as just about any engineer will know what I just said. What I think it's about is driving the headlines, "Tesla engineers are doing a code review at twitter" sounds very important to the lay person.
Twitter has been publishing about their architecture and all the updates to it for years and years, and given countless presentations on it. https://blog.twitter.com/engineering/
My guess is that at least part of the reason to bring in these devs, is to make sure none of the code or git histories (or their backups) are tampered with.
Many twitter staffers seem willing and motivated to sabatouge the new owner over purely political issues. A few have even suggested it on their own twitter accounts. Elon is polarizing and that leads to both more challenges and more opprotunities.
There's a big difference in this case between a position of authority and an authoritarian leader.
Typically most people would regard good leaders as people you follow because they inspire confidence, you feel like you can trust them, and that your efforts will be rewarded. Authoritarian leaders on the other hand are typically those who use their ability to punish people to force compliance.
While the owner of a private company doesn't need the approval of their employees to make decisions I would never work for someone who uses threats and doesn't explain the rationale behind their decisions.
Perhaps a better analogy is that the CEO is captain of the ship. You get credit for things going well but if things go wrong you are responsible. It does not matter who screwed up. There's a wide range of ways to implement that model ranging from despots like Chainsaw Al to consensus builders like Jim Whitehurst. [0, 1]
"I dunno why any Twitter engineers would be really trying that hard right now.
Passively destroying Twitter from within by not trying, and trying to make a billionare lose 44 billion dollars is a significantly more fun and interesting challenge than making Twitter a better product."
>“Fifth column” paranoia is a hallmark of authoritarian leadership.
Is it paranoia when within this very same thread people are making posts about ways to see the company fail? By omission or direct action, if the ideas are expressed here then they are held by workers at Twitter.
Do you fail to see how this is relevant?
They comment I replied to is making a rather poorly founded insinuation of authoritarian leadership, ie Musk, despite having 'lead' over the enterprise for days, at best.
Presently this jibe, 'authoritarian' is being levied at conservatives & most absurdly at free speech advocates. The idiocy of such an assessment is, in fact, the operation of an authoritarian ideology.
It’s really hard for me to imagine myself working at a company that deals almost exclusively with machine-generated data (as Tesla does) and being assigned to judge a social platform’s engineering qualify in one day. Cars and spaceships on a closed network are fundamentally “clean” compared to having hundreds of millions of people who produce dirty and often adversarial data.
But I guess the point of this exercise isn’t really to judge code quality, but to drive fear at Twitter.