cowanon77

I dislike this whole genre of “everybody is doing x wrong”. The argument almost always boils down to “I’m super smart and everyone else is an idiot”. Maybe other people just have different needs and preferences?

The post doesn’t mention a year. Early Microsoft Windows was very badly documented and the docs were difficult to search, so it wouldn't surprise me that it took senior level support to know that fact.

I think it's a good thing. In a couple of years we'll have a clear way to measure the true productivity and impact (or lack thereof) of AI on software dev. It's the perfect "natural experiment".

The dislike of giant data centers is also authentic though. They are generally bad neighbors, for a wide variety of reasons (water/electricity usage, loud, over lit, generally cheap construction with no aesthetics). There are ways to be more neighborhood friendly - you often drive by traditional data centers without even knowing.

You just can't extrapolate much from a study of this size, for this effect size, with this many confounding factors. The difference in effect sizes were percentages, not orders of magnitude.

The best you can say is: "no evidence found of non-efficacy for these 3 diets. Effect sizes comparable to within 10% of each other."

There's no way to interpret this as "x diet is the best"

This may be true, but the study size is 42 people. This is woefully insufficient for making any conclusions about something with so many confounding factors. At best this is a safety scan before running a much larger trial.

Yeah, I disagree with SVP's approach in this story. First, get the current issue understood, resolved and remedied. Then, thank the team for their good work in fixing this issue, and do a post-mortem to see what process improvement could have prevented the issue from rising in the first place, or at least detected and mitigated it more quickly.

I only see 2 consistent world views: either intellectual property is real, or it is a false concept and all information should be free.

If IP is real, then the AI companies have performed flagrant theft.

If IP is not real, then the algorithms and weights the AI companies have developed should also be free as they are just more information.

The status quo of "your knowledge has no protection, but our knowledge is sacred" is the worst of all possible worlds.

I do have compassion for people who are not tech oriented. AI art can seems a bit kitsch and novel at first, and they probably don't really think about ethical consequences very deeply.

The problem is it all looks the same, and those of us who are terminally online are exposed to it non-stop. AI slop seems to be poisoning every space that used to have quality conversations, and mass un-employing many creative professions.

People do have the right to vote with their feet though - it is perfectly ethical to support coffee shops that support the local artist community and avoid ones with AI "art". Too bad we've mostly lost the ability to rationally discuss things, instead of resorting to threats right off the bat.

AI does this really well, even better than new development. I regularly use Github copilot to automatically read production logs from the app and backend services, find errors logs, deeply scan the codebases, create theories about the issue, verify the theory matches production evidence, verify the theory in a sandbox env, create a failing test, and then suggest a minimal code change in a feature branch, verify test is passing now in sandbox, and submit a pr, along with a detailed analysis doc and any future design optimizations to fix the class of error.

AI is also very good at scanning legacy codebases, and reverse engineering design docs, system interfaces, and trying to reverse engineer what requirements and design tradeoffs lead to that system. It's not perfect, but already at medium software engineer level.

> we are passing in training data that says to do those felonies.

Partially, but also I don't think current AIs really have any judgement of right and wrong, they just see chains of reasoning between ideas. This is the deeper issue, there is no way to sanitize the data or training to fix it. Current AIs are fundamentally unsafe, and only become more unsafe as they become more powerful.

We still have some product development software teams at my company that have not changed at all in the last year. We develop medical products in the U.S., so it is heavily regulated and processes change very slowly. We also require reviews from our legal team, so it takes a while for any new contract or agreement to be approved.

This is temporary though; other teams are fully embracing AI in their processes, and it's really only a matter of time until all teams use it. We have C-suite instructions to promote AI adoption. AI will slowly creep into software testing, verification, and eventually the requirements management and software development itself. I'm working on some of the pilot projects.

Personally I have found "meta-prompting" to be much more useful than any pre-canned set of skills. Simply ask the AI (inside of a workspace already set up):

Create a standalone prompt to <xyz>

The latest AIs will print out a long prompt with all of the assumptions, tools, and general files it plans to use. Review that, and then run the whole prompt in a new context.

I used to really enjoy writing code, to the point where I would have thought of it as art and fine contemplation.

25 years of corporate development beat that enjoyment out of me. AI is just the final nail in the coffin. I have no enjoyment left in coding; I firmly believe the current crop of programming languages will go in the same vein as human written assembly language did.

I don't believe the current LLM-based AI will replace most other jobs, but I think it is well suited to completely replace much of programming/IT and morph it into an unrecognizable job role.

This the the heart of the problem.

The biggest problem on these projects is purely human - no stakeholder wants to inconvenience themselves for a vague bigger picture benefit. The executives don't want to get involved in the weeds to make decisions. On top of it, modern business operations is complex and no single project manager or architect understands all of the concerns and tradeoffs. So the whole project devolves into massive bureaucracy and the smallest decision takes a 5 or 10 person conference call, and the result will often later have to be justified to everyone not involved.

On top of that, the best talent often avoids these projects because it's often tedious and boring work with little individual reward other than the salary.

On the surface it seems bad, although many of these books were probably rotting in place rather than being read. If the companies are willing to make the digital version available this may actually prreserve the books.

More troubling to me is the focus on pre-2022 data. This indicates that the companies are seriously worried about model collapse, where the future models become worse becuase they are being fed by generated data instead of real human data.

This is actually my worse case scenario for AI: the models become good enough to disrupt entire industries in the next 10 years, but then stay frozen at that level. Model drift then kicks in because the world will keep changing but the models do not, and in a few decades we actually regress instead of progress becuase they won't be enough skilled humans to drive advances.

It seems to be true this time though; I have observed it myself and heard it from several experienced developers I personally know and respect. It feels like some threshold was crossed with Opus 4.5 and Gpt 5.3, where the models are now able to reliably solve certain classes of problems that were previously unreliable.

Time will tell of course, and it’s early, but inflection points do exist with progress.

I’m not sure why that’s controversial - I have met many Ivy League students and grads; they are all intelligent, at least in an academic way. The only other common characteristic is that they almost always had some form of privilege. Either rich parents, or adults around them who worked very hard to get them to that level.

A lot of upper middle class people recognize that AI is a direct assault on their livelihood. The very jobs that AI threatens to disrupt are the bread and butter of the upper middle class.

Companies are fundamentally untrustworthy in the long term. You are always one management change away from the product being unusable or hostile to your use case. With old fashioned technology there is almost always a path to fixing it long after the company is dead.

The problem is computer technology in its current state is fundamentally hard to trust. Without reading the source code and knowing the full source of all external services, and hoping the terms of service or external source code don’t change in the future, you really can’t trust anything. There is no authority that can guarantee “this will always work until it physically breaks, and even then be repairable”. Conventional parts and circuits can be more easily repaired and even reverse engineered if needed.

Jira is only a set of changes though. What happens on a long (10+ year) and complex (10+) developer project with many changes and revisions? Eventually you need an explicit specification that itself has a "current state", and a change log. Theoretically you could generate this from Jira, but in my experience it eventually became a mess on any larger project that didn't have explicit and maintained writen requirements.

There is an imbalance in leverage and timing though. Dynamic pricing requires a lot of real time and historical data; companies can access and share that information easily, and you as a consumer cannot.

Even in areas with multiple competitors, they can (and do) effectively collude by getting their information through data brokers and third parties.

I don't have a solution, but we are currently very far away from a free market in general.

The core theories of economics seem correct, and in fact predict the current situation if used correctly. The problem is that the game has been systemicaly rigged. Employers have maximimized their legal leverage, while crushing all employee and subcontractor leverage. The central banks favors (and often rigs) the metrics that favor the rich (GDP, U3, CPI, etc), and ignore alteratives that favor labour. The government designs tax policy to favor integenerational wealth, and capital accumulation over income.

The new deal in America roughly got things correct, and was followed by the greatest expansion of the middle class in history. What we're suffering from today is the systematic destruction of that social contract.

It's similar to the problems with GDP, CPI, U-3 (Unemployment rate), etc. You really need at least 30-100 metrics or more to get a grasp of what's really happening in a complex eonomy. Debt can be good if it causes more growth in the economy, and bad if just gets trapped in a few foreign investment funds. But the number by itself doesn't tell us any of that, yet that is all that grabs the focus.

Often politicians will try to shrink or control the debt without deeply investiagting the true causes, because the real causes may be politically inconvenient.

This is the big misnomer - it's debt on the U.S. balance sheet, but is a credit on someone else's balance sheet. By itself, it's not necessarily bad; it really depends on who the debt is issued to, why, and on what terms.

Time would be better spent talking about other issues, but national debt is a simple number politicians can complain about rather than talking about the more complex issues that really impact the rest of us.