I think if you read a little bit further in the article you would find that this attitude is exactly why the author believes there is a collapse in non-fiction book.
> The collapse of nonfiction is [...] a massive breakdown in the social fabric of trusting education, expertise, or any authority
People don't want to read non-fiction because they don't trust the people writing it.
There is a famous story about how an F16 got grounded because an half inch metal pin was missing from the cockpit canopy. Nobody knew what the pin did or if it was even necessary but you can't fly a plan that 'missing a part'. After several days of being grounded they finally reached out to a guy who had retired 20 years ago who told them the pin was a part of a analogue sensor that had been replaced nearly 30 years ago. Only the pin remained as a vestigial organ since it was part of the airframe design.
It doesn't matter how well you document things there will always be tribal / personal knowledge.
Even if stuff is written down you have to know where to look.
I babysat for a military fighter jet mechanic when I was young, he said when they dismantled and rebuilt planes there was a drawer for all the spare parts that were left over.
It’s a great story, but I assure you there is a pretty high percentage of airplanes flying with missing parts and systems placarded as INOP in both military and civilian aviation.
Before you can placard something as INOP, you must determine whether the manufacturer and certifying authority have included that part in the Minimum Equipment List or Dispatch Deviation List. Mechanics do both get to determine this at their own discretion. Hence the call to the manufacturer for clarification on the necessity of said pin.
To reiterate, you CANNOT fly an airplane with an undocumented missing part. If the part is documented in the logbook as removed, and it is acceptable to fly that way, then it is not missing.
I have been building and maintaining commercial aircraft for a living for 30 years. What is your experience?
Thanks for asking for my credentials, even though you don’t seem to disagree that lots of aircraft are dispatched with imperfections, so many in fact that there’s a defined process and CFRs to cover it.
Because you asked, I’ve been flying, owning, and overseeing the maintenance of our aircraft for 27 years, including understanding my freedoms and obligations as a pilot and as an operator with regards to the legality of dispatch for imperfect aircraft under both normal operations and under special flight (aka “ferry”) permits. I have spent a lot of time across those three decades both socially and in continuing education courses/seminars with both GA and commercial flight crews and A&P/IAs.
I don't think you guys are really disagreeing with each other. Just that one is using a colloquial definition of "missing" and the other, a technical one.
The 'forest for the tree' here is that when it comes to general operations, a crew chief doesnt generally have the authority to signoff a bird as flight-ready for an unaccounted component. That is the authority of a PMO or a Risk Officer or someone else in flightoos, not a frontline maintenance call. Maybe someone in a maintainence squadron?
An absent (removed) part might be acceptable depending on circumstances, but a genuinely missing part creates hazards and must be ascertained as to its location. A mechanical jam, an electric short, a hydraulic leak, or a Foreign Object Damage to the engine from a piece of metal on the loose can destroy the plane, endanger people and mission. Same concern with tools & bits, can't have any unaccounted for.
FOD applies more broadly than that, as well. Candy wrappers on the runway is also FOD.
It stands for foreign object debris, and means any physical thing out of place where it may cause damage or injury. Because of the nature of jet engines, basically anything can be FOD depending on where you put it.
Exactly. Framing completely changes this from an absurd story causing pointless grounding due to loss of institutional knowledge, to an example of good judgement (or protocol) in the absence of knowledge and acknowledgement of the risk of unknown consequences.
What's being ignored is the 1/100 of these stories where this judgement pays off, saves aircraft and lives that make the other 99 worth while.
> Even if stuff is written down you have to know where to look.
That's one of the things I love about AI: It's excellent at figuring out where to look, because it can digest so many documents so much faster than a person.
Yeah it’s great to point an AI at a very detailed and specific engineering manual and have it distilled down into bullet points. There’s certainly no room for information loss in that plan! /s
Have you seen the maintenance manuals on an aircraft production line? They're in 3-ring binders, so you can take out and replace pages. They are constantly being corrected and updated. The industry recognizes they're never at 100% (and to their credit mitigation is part of the process).
And because it's accurate the AI will know how to weigh that accurate information against other accurate information in such a way that it will give you a great answer without any help. In other words: any amateur will be able to do aircraft maintenance! What can go wrong? /s
GP did not suggest anything of the kind, though. They talked about "figuring out where to look" which very different. In that kind of use, it doesn't really matter that the model might hallucinate or draw faulty conclusions, as you're not using it figure out what to do, how or why. The only thing it is telling you is where some information is located, after which you go check it out from the actual source of truth. If the reference turns out to be wrong, you'll notice immediately.
I think you are broadly correct, but just to pushback on a few points:
(1) Ability to solve hard problems in days vs weeks as immense value
(2) Back-end improvements (if done right), should improve platform speed, stability, scalability etc. which should have revenue implication
(3) Ability to on-board a SWE equivalent entity in minutes, have them work on a specific hard problem and then off-board them in minutes can have value
All of the above, of course, depends upon Fable consistently being a 2x-3x SWE at minimum.
You're not really solving problems, you're retrieving the best match of solved problems from compressed corpus. And that corpus is available to many companies, meaning "hard" problems stop having "hard problem" value the moment they enter the weights of any model via the internet ... or distill from one model to another. Anthropics business model is commoditising knowledge, but as we see with the Fable model card, they only want it done to the knowledge of other businesses, in their own field, they totally hate it.
I don’t think that’s an accurate or useful characterization of modern AI like Claude at all. It is not simply regurgitating knowledge. It applies its knowledge to create bespoke solutions to the problem you pose to it, and is able to self evaluate its progress towards the completion criteria. If you don’t think that counts as “problem solving”, your definition would exclude nearly all knowledge work and engineering.
People underestimate the vastness of training data (internet) and overestimate their ability to recognize if something is really bespoke. Not to say the no problem solving is happening, because there are many problems that we inefficiently solve again and again and the LLMs are making the solutions more accessible to everyone with a subscription.
> It applies its knowledge to create bespoke solutions to the problem you pose to it, and is able to self evaluate its progress towards the completion criteria.
It imitates applying knowledge. The imitation may be uncanny, but assigning LLMs intentionality and ToM is a category error.
Does "applying knowledge" necessitate human-like intentionality and theory of mind? If you insist it does, and this is a category error, then we need a new category.
By analogy, consider that many have referred to classical, deterministic computing as some kind of "thinking" for the last half century+. Does this stop being kosher when the computer has an uncanny propensity for human language? Perhaps, but the computer is still clearly chewing through problems that would have required a lot of human thinking (e.g., arithmetic) in ages past.
I haven't seen any genuine proposals for words to replace the human mind analogues, let alone proposals that the anglosphere would plausibly adopt en masse.
> You're not really solving problems, you're retrieving the best match of solved problems from compressed corpus.
This is not correct. LLMs interpolate in a high dimensional space, so you're actually composing the best matches in a compressed corpus to find novel points/paths in that space. That is problem solving.
The thing about AI-generated “solutions” is that they often go down bad rabbit holes and need to be re-run, or since they are so “cheap” to create they are often just thrown away and rebuilt when requirements evolve. Plus, just more stuff is created and needs to be maintained. So in the end, your efficiency gains go out the window.
20x the cost means you need to have fable to be 20x better than the alternative, which is a tall order. And there's more options out there too, perhaps the 4x cost is enough.
This means if the deepseek / under 1k alternative is at least x1.2 improvement, fable needs to be x24, which I think is very2 unreasonable. It is possible for it to worth if it can x2 a $20k SWE, though I doubt it can do that.
“Ability to solve hard problems in days vs weeks as immense value”. Citation needed.
LlMs are incredible don’t get me wrong, but they are good on tiny contexts (writing a script). Not on software engineering (adding features to Chrome).
Honestly, LLMs been OK at adding features to software since around Opus 4.5. From what I've tried of Fable, it's a decent step up from the Opus models and I can only see things getting better.
Look I don't like Tesla as much as the next person, I think it is wildly over-hyped and over-valued. But this article is just slop.
The headline says - "How Tesla hid accidents to test its Autopilot" but the actual article has no explanation as to (1) how Tesla hid anything or, for that matter, (2) who did Tesla hide this information from
It mashes together a Tesla data leak from 2022 and an unconnected lawsuit from 2026 without ever explaining how those 2 are connected.
Tesla has a pattern of making deceptive promises and deceptive disclosures but this article doesn't make that case at all.
>Tesla has a pattern of making deceptive promises and deceptive disclosures but this article doesn't make that case at all.
This is something I find frequently as well, moreso with Musk related things than Tesla. Lord knows there are plenty of things to be critical of.
If investigative journalism wants to regain the respect it once had, fewer allegations with concrete claims serves both the public and faith in media over large quantities of vague claims.
I admit if you want to sway public opinion, the latter is more effective, but is also a mechanism that doesn't require alignment with the truth. When that approach is normalised, it opens the door for anyone to shove popular opinion around.
After you wrote this, I went and read the article I also didn't see much there either. And wonder why you are getting down voted. And TBC, also not a tesla fan (the truck is dumb).
Coca Cola indeed has a lot of value its market capitalization is USD 76 billion making it one of the 30 most valuable companies in the world!
The problem with SpaceX is that its valuation is almost entirely driven by its expected future growth. For 2025 SpaceX reported EBITDA of USD 7.5 billion. Other mature aerospace companies (Lockheed, Northman, Airbus, Boeing etc.) are currently valued as ~19x EBITDA (i.e., Market Cap / EBITDA is ~19x). But SpaceX is being valued at 166x EBITDA (USD 1.25 trillion market cap / USD 7.5 billion EBITDA).
What drives this difference in valuation? The answer is quite simple, investors expect the EBITDA to grow and quite rapidly. EBITDA could grow via higher margins (EBITDA margins is EBITDA / Revenue, and for SpaceX it is already a decent 50%), but even at 100% EBITDA margin (i.e., zero operating cost) its valuation multiple would b 83x EBITDA. So the only way to justify SpaceX valuation is if its revenue grows and gorws rapidly.
A quick back of the envelop calculation would shows that investors expect SpaceX revenue to grow at minimum of 65-70% annually for the next 5 years. If the revenue grows at less than that the investors are unlikely to earn a good return on their investment.
Someone else here pointed out that when your biggest asset is a network of thousands of satellites that all have a five-year lifespan, earnings after depreciation is unusually important.
> The collapse of nonfiction is [...] a massive breakdown in the social fabric of trusting education, expertise, or any authority
People don't want to read non-fiction because they don't trust the people writing it.
reply