heisenbit

I‘m really wondering about plan and then agent mode switch in Copilot even with the same model. The switch swaps the prompt so all the learnings during planning which are in the cache will get invalidated.

It was also the rise of Github and the importance of the software hosted there. More consistent documentation and transparent issue trackers/PRs helped a lot dealing with evolving software.

If we are in the long run stabilising that would be already a win. Until the temperatures go down it will be a long, long, long time if ever. So temperatures will go up and we have to adjust. The scale of required adjustments is not understood yet and the increasing number of surprising needs to adjust is scary.

Plus the power needed for cooling adding maybe 50%.

Not necessarily. By handling certain aspects orthogonal to the main flow it reduces the context the llm has to keep track of and should enable deeper reasoning of the main functional logic.

Reproducibly through measurements has a lot of value if you do not have a coach like the author for his mother. Yes a lot of recipes can be handed down but the space of recipes is so much larger. We don‘t need books as we can tell stories and this bring out the feel so much better is true but where would we be without the encyclopedia.

In December by chance I put a pack of Vitamin D into my shopping basket. I did not think much, thought to take 1000IE but then decided that for the first week I take 3000 to catch up. Muscle pain went and control over eating improved. I did not expect any changes based on past experience with 1000 but this time I could not ignore it (age can play a role) and I stayed on 3000. Tests a month later showed I was just not deficient any-more. I continued on the regime and started having improvements in long running skin issues to the extent my dentist noticed. It may not be a miracle drug but one should not underestimate cumulative impact individual factors, age and lifestyle changes (less sun) that may change levels and demand.

Changing abstraction is a high risk unlike agents refactoring scores of almost identical code.

Looking at actual combat robots in Ukraine Boston Dynamics robots are conceptually different and not necessarily in a convincing manner.

You forgot to mention that all the people who made the original product great left and all the ones which could make the successor great did not join.

More like: All the rules that determined how the internet services and content were financed are out of the window and are rewritten by AI.

Ripping out sensors is the equivalent of shooting the messenger.

Some of the best books on JS which were online went recently off-line for that reason. Blog post by the author: https://2ality.com/ (Dr. Axel Rauschmayer)

I think it is work to set up but I'm also learning a lot setting it up. Mainly using qwen/qwen3.6-35b-a3b mlx with my 48GB M4 MBP which leaves me just enough headroom for docker dev-container and other basics. I use LM Studio to run and am using it via VSCode. A big difference made the system prompt improving the tool integration (I asked GPT for guidance on that). Before that it was not making changes but regenerating code often messing up than helping.

I mostly run my MBP on low power even when it is plugged in to avoid the noise and heat. Full power maybe doubles speed but more than doubles power.

What can it do: Simple restructuring of pages. Where did it and other models fail: Splitting up Pinia store which GPT-5.4 did without fail. I think with more tuning, guidance for tool use and maybe some support tooling around it performance can increase further.

Isn't Grok part of the IPO?

I just tried Google search in Germany on my iPhone: AI results AND the disclaimer was behind a „show more“ button i.e. the may not be any disclaimer (and when shown it was in a small font).

And the more you protest the more your name will be associated with child trafficking. Streisand effect multiplied by LLMs being not good in dealing with negative information.

Now if they just let me switch off the sound when I connect the charger. For any couple not going to bed at the same time and charging their phone at the bed this may be a welcome innovation. I'm willing to license this idea for free.

Limited liability makes taking unlimited risks a rational choice. AI ‚only‘ scales this corporate model up and compresses the timeframe to the next disaster.

The article vastly underestimates how often no was said before ZIRP. Getting to yes was very, very, very hard. Zirp moved the default to build it snd they will come.

It is so comforting to deal with known unknowns particularly when the unknown unknowns are the ones that get you.

How would one call such a strategy? Embrace and extend comes to mind.

We always knew the limits to Moore‘s Law are first and foremost economic. Given an industry used over decades to predictable lowering of price per compute function and thus swallowed any advance for new user functions and overhead when the limits are reached there is going to be a squeeze. AI scaled up at the time the production capacity became more inelastic.

Maybe it is time not just shrink transistors but also software bundles. I can see decades of possible progress hiding in plain sight behind a browser screen.

One value of learning on my Macbook is that mps is not as well supported as cuda which forces me to go down roads I would not have traveled.

every keystroke, with millisecond timestamps

Finally someone has figured out how to hire accurately.

There is a reason moderation decisions are not perfectly transparent: They are gamed otherwise. So there needs to be legal recourse with discovery and meaningful liability attached to submitting to the role of acting as the agent of a foreign government.

While I agree with adding code contributing to complexity is problematic there is lots of code in existing code basis which is overly complex due to past outdated requirements or less than perfect human coders. The current flood of AI driven security fixes demonstrates that AI can be pretty good in detecting security edge cases. It is not inconceivable to use it to also reduce code complexity.

And they are enticing human agents to further their agendas using techniques learned from the white mice.

Yesterday in Lidl I was a bit shocked seeing the coupons offered by their app. They did a really good job with their mixture of stuff I had bought or might buy.

My suspicion is that it violates the users copyright on their commit message.