bob1029

I've been considering very unconventional techniques for interviews. Like requiring candidates participate in a team based multiplayer game as part of the interview process. Completely out of left field without warning.

Assuming the applicant has zero experience with gaming, this would be an incredible opportunity to evaluate how well they cope with uncertainty under pressure and adapt to unusual conditions. If the applicant is already a grandmaster on the ladder, this becomes an excellent test of their humility and teamwork.

If the applicant gets shitty about the mere idea of playing a game as part of a "serious" job interview, I think I would have all the information I need to make the right decision. If you can't get a little bit flexible in this new world, I don't want you on my team anymore.

30 years ago we caught them in glass jars. They were so dense in rural Missouri that you could catch several at a time by swinging an open container through the air.

Don't forget about the arachnids.

Spiders are like senior members of nature's janitorial staff. I've seen them convert a periodic trail of ants into a mass grave within a few days.

Everyone I know who hires an exterminator has significantly worse issues than the occasional spider sighting. I don't actively encourage them but I also don't try to clean their homes out aggressively. I'll do yard work around them (they get until next time to move). Arachnids are unusually intelligent given their size and often seem to get the hint.

I am at a point where I would actually prefer to be the critic than the author.

If you told me my only job was to review PRs created by something like gpt6.1 on xhigh, I would have very little problem with this. As long as the upstream issues are kicked off by a human who is ultimately responsible for the outcome.

PR reviews can go very quickly if you get out of the ideology of owning every single line of code and begin focusing more attention on the data, schema and other structural elements of the solution.

The frontier of AI has gotten so spicy that the delivery risk of pursuing this rabbit is astronomical. Knowing when to quit and use someone else's solution is critical at this scale.

If someone started from zero, how long do we think it would take for them to reach the cost/capability front currently provided by gpt-6.1-terra? How long until OAI figures out a way to distill that down again for pennies on the dollar?

Apple seems to have this sorted. You won't find them chasing OAI, Google and Anthropic around the playground.

https://blog.google/company-news/inside-google/company-annou...

Just do the schema by hand. It forces so much downstream alignment that you can almost prefer to use the schema and any associated comments over prompting the model directly.

If you don't know what your schema should be, then I think this is the dead canary in the vibe coding coal mine. Back up to the design phase and talk to the robot about the domain until you have a really strong conviction about its tuples and relations.

You can blame reckless pursuit of multithreading for a lot of this. Most UIs should operate using only one thread. Communication between threads is unbelievably expensive if you don't actually need to do it. The multithreading thing turned into multiple computers and containers. It's all an extension of the same madness.

Until you can demonstrate to me that you can build an async/await server that uses a single core without deadlocking, I am not so sure you should be trusted with any additional resources.

You could take this further and argue that Diablo 4 further "solves" the problems of Diablo 3. It is accessible to a far wider audience than D2 and D3.

For someone who played D2 and D3 for thousands of hours, D4 typically comes off as really awful. The accessibility tradeoff means that the cheesy (and enjoyable) strategies you used to enjoy in former iterations are no longer feasible because the auto adapting numerical difficulty system is so aggressively casual. Every time you get a big piece of gear, it is watered down to pointlessness immediately. D3 and D2 do not have this problem, but their difficulty curves are more discrete and unforgiving in many areas.

> let me tell you about health care coverage

I haven't had any coverage in 3 years. It's not the end of the world. People who lose their shit over weird tail risks are missing out on the competitive advantage. You can compensate for the expensive doctors and prescription meds with more exercise and cleaner diets.

I always looked at employer sponsored health care as a sort of trap. I never got involved in it. Every insurance premium I would have paid has gone to a retirement, investment or savings account instead. One big open runway is a lot more useful than a bunch of smaller ones with weird constraints. The idea of a health savings account amuses me.

As long as grid operators keep an eye on their inertia figures and fault current handling this is great news.

Synthetic inertia created by inverters can maintain grid frequency, but they cannot handle fault current. Without the ability to ride through disturbances, downstream protection equipment will fail to work correctly. If you hit a semiconductor with 10x its rated load, it needs to shut down immediately or it will be destroyed. The turbine at a nuclear or coal plant can take a serious beating for several cycles before it would be forced to trip offline. The machine is massive enough to absorb any damage without losing its overall functionality.

We need actual synchronous machines spinning with heavy masses. They don't need a carbon-emitting prime mover on the other side. They just need to be connected to the grid. They also need to be placed relatively close to the inverters (same plant/substation) or they won't provide the desired support.

I have learned that merely thinking about deadlifting a personal record can produce more adrenal response than the act itself. I think it might be better to draw the mind trip out a little bit before you do the lift each time. I feel like it reduces the chance of injury when you give your body a few moments to prepare for the onslaught.

I was thinking about the possibility of getting an international phone number and using it domestically. The spam is really out of control in my area.

The results for factual error rate on "difficult" prompts is staggering to me. Particularly the difference between 5.6-Terra and 6.1-Terra. These are wildly different Pareto fronts. Our custom enterprise agent is currently running on 5.6-Terra and it's already very good. I look forward to updating the configuration.

> For example, when we served Kimi K2.6 with vLLM version 0.28, we got the expected tool call schema only 92% of the time. In other words, 2 out of every 25 attempted tool calls failed.

100% success is critical for tool calling.

The probability that the overall agent iteration is successful would scale very poorly even if we achieved 95%. With p95, if you need 50 tool calls that must execute sequentially to achieve a result, the chances that you won't have an error are only ~8%. There are ways to recover from bad tool calls (error feedback helps a lot), but if the agent is persistently using the wrong tools with the wrong args the overall system will not work very well.

I have a hard time believing the camera bump isn't the result of intensive consumer psychology experiments. Setting a bare apple device down on a flat surface feels like chair with a leg sawed off. They need you to buy the accessories. This is part of making the ecosystem viable.

Neuroevolution is the most promising and the most cursed form of ANN technology.

Evolving the topology and number of parameters makes interpretability nearly impossible. But, this is also what provides the opportunity to achieve incredible performance with orders of magnitude fewer parameters.

Review the specifications for the latest Telum II chip if you need a hint. 5.5ghz base clock.

If you absolutely need to scale something vertically, IBM Z is the final stop on the line.

X86 is great at hosting an entire farm of applications. You would never want to install salesforce or gitlab on a mainframe. That would completely defeat its purpose. If you've never encountered an appropriate workload, IBM might never make intuitive sense.

Also, the roadmap:

https://newsroom.ibm.com/2026-08-24-ibm-unveils-next-generat...

> Second, the semiautomatic washing machines themselves may have taken some of the drudgery out of clothes washing, but they did not really save time. Scott explains that the person washing “had to manually feed each item of clothing through the powered wringers between each part of the cycle.” You could not just close the door and turn the machine on and leave.

Second? Why not first? This could be the entire article. Why would I buy a machine that doesn't actually provide meaningful value? Everything else in this article is ragebait for HN.

I've tried a few things that I know are possible (multiplayer FPS with client-side prediction), but the overall challenge still persists. There is no easy way to rapidly iterate this architecture with an agent unless you already have massively deep pools of experience to draw from. Launching this system into the real world is even more challenging than building it.

For many problems, the LLMs look like they might be making progress, but once you zoom out far enough you realize it's just pushing the toothpaste around in the same tube with the cap still on. The mass of the overall system is not changing despite the appearance of localized progress.

If your primary source of identity was gatekeeping the Pantone colors of the various office furniture, then yes perhaps the challenge is now gone. But, if you were focused on mission and 'big picture' (what furniture and why), it's plausible that the challenge could actually increase in its intensity.

> Naturally, game engines are crazy optimized, so the UI runs butter smooth and uses pretty much no resources.

We should build everything using a AAA FPS engine loop now that we can. The amount of shit that happens in one rendered frame of overwatch is wild. And then we go and do like 200 of those per second.

Our processes around software should come first.

A lot of the AI discussion seems to be focused on what these tools mean for an individual developer, but in a team setting any gains are at risk due to coordination tax and ideological friction. In my experience the ideological friction is most of it right now.

Figuring out a way to accelerate the process in light of the code writing bottleneck being removed is essential. I've observed a development team regress back to fully manual development because their daily PR/code review process is too slow. It doesn't matter how fast you write code if you can't actually merge it and ship it to the customer faster too. PR stacks can only take you so far before the conflict monster gets too big to fight.

I don't think AI can fix the process, ideology or people issues. I've seen how long an average PR review actually takes. If these employees were simply doing their jobs at all we'd be 10x better off regardless of AI. A truly skilled developer who makes a genuine attempt should be expected to solve 99% of their tasks within the same business day as of October 2026.

Burning down the backlog to zero doesn't mean the development team will be subsequently laid off. Quite the contrary. It will create a vacuum that the business will fill with new work that may incidentally increase the developers' compensation packages. No one is laying off the development team that just chewed through 50 items per day for 3 straight weeks. Who would support all the potential bugs they just created? That's right. They might hire even more developers. This is called a "win-win" scenario. Not every game must be zero sum, but I understand rejection of this possibility to be part of the aforementioned ideological issues.

I am the same way with MSSQL. I used to avoid it like the plague because of how the permissions/logins/everything worked. Now I can throw a T-SQL script at it in a few seconds to make it behave.

I think we are struggling to achieve payoff within the tech sector.

There have been wins with the customer facing side of the business, but the development team seems to be the biggest problem right now despite the direct & obvious applicability of these tools. I know they aren't a special unicorn edge case because I was very rapidly solving their issues before they stopped handing me things that revealed the lethargy.

Within the sector where these tools definitely work, the biggest barrier to effective adoption is figuring out a way to undermine the various ideological high grounds. I have found the most effective argument to untangle this knot to be something approximating:

Challenge:

"Why should I adopt or help to implement technology that might eventually replace my job?"

Response:

"Because refusing to won't protect your job. It puts everyone's job at risk."

It doesn't matter if you protect your job (not using AI) if the entire company goes to the bottom of the ocean as a consequence. The neat thing about this is that your job is part of the set known as "The Company" and if that ceases to exist, so does your job. Good luck suing a dissolved entity for unemployment or whatever. There is enough competition in most markets to punish anyone who thinks they can win on virtue alone.

(Assuming the move was due to the recent outages)

I don't know that the grass will remain green for very long anywhere in this current scenario.

GitHub is not having most of its outages because GitHub's engineers are using AI (although there might be a case or two of this). The primary driver of the outages is the fact that developers can move so much faster in their workflows. Every single user of these products is placing a heavier demand on the system. I don't think this ramping of demand will cease any time soon. I fully expect to receive an email any second now about how my 5,000 req/hr GH API quota will be cut in half before 2027.

I don't understand how GitLab would be immune from these forces, unless we were willing to reach for self-hosted & enterprise plans. Both of which GitHub offers.

I run a Midea dehumidifier around the clock during the summer. The inverter driven compressor noise is quite pleasant. It masks higher pitch tinnitus very well.

It's probably something around the idea of an external oscillator influencing the one in our heads.

https://en.wikipedia.org/wiki/Injection_locking

> move keys between HSMs

This should rarely happen if we are doing things correctly.

The moment the private key leaves the secure context of the HSM you should consider it to be compromised. Some HSM vendors have a way to effectively extend the secure context across multiple physical machines to address the edge cases where you'd be inclined to do this.

> The external consequences are usually immediate bankruptcy of the CA after the investigation concludes.

It would seem that we have different definitions of "wind-down". Everyone in the business operated with this consequence already on the table and it did not seem to provide the desired outcome.

Accountability is rarely something that a person or organization will spontaneously adopt of their own free will. It generally has to be forced from the outside by way of additional consequences.

Right now, it would appear that the strongest external consequences for fucking up a CA are strongly-worded posts in a web forum. I would hesitate to call the wind-down of a business unit a "consequence" in this context because it's the default outcome. Everyone was already playing that game. We need an additional layer of pressure to encourage good behavior.

I experienced this problem in semiconductor manufacturing.

It was really difficult to explain to Korea that the additional warnings they wanted me to put in front of the tool operators were actually causing more harm than good. The cultural divide made this really difficult to approach.

"What do you mean the American mfg staff don't actually read everything?!"

"Yeah... Especially at the ATX line. We kind of go off vague shapes around here"