ComputerGuru

Would you be able to run it against Gemini Flash (not Lite) 3.0, high thinking?

That’s only for tool use.

Do infra providers reveal that level of implementation detail?

Supposedly there are “jailbreaks” that expose considerably more of the thinking traces.

Did I miss the link to the study? I was wondering if storage contamination were a possibility? Breast milk storage bags are all plastic, and cheap brands abound.

On-site only?

I would have liked to see a bit more on the theory side of things, explaining optimal weight and inference splits, actual issues with existing drivers, etc instead of what’s essentially just a recipe.

The problem with trust is that it is easy to lose and hard to get back.

You can't blame the people commenting "they SAY they won't silently sabotage your session but how can we know?" because they're right, we can't ever know. And Anthropic has firmly planted the seeds of doubt.

It doesn’t even remotely need to fit on one SSD with logical volume management (or RAID).

No, the main difference is that MySQL bundles the code needed to interact with the old db version in the newer server binaries (effectively by not changing the on-disk binary format!) while pg_upgrade requires you to have both old and new installs living side-by-side to reuse logic/code from old binaries. It is a more bulletproof method and less susceptible to bugs and (upstream) developer errors, but is (or at least can be) harder for the sysadmin+dbadmin.

(For example, ports under FreeBSD doesn’t let you install multiple Postgres versions as they are marked as conflicting packages so installing one necessarily uninstalls the other. The saving grace here is that most (virtually all) FreeBSD installations have root on ZFS and you can employ ZFS snapshots (via the hidden .zfs folder) to access the old binaries after upgrading to the new postgres version, but not many people know this trick!)

Thanks, I missed that on first glance and did manual translation.

Not sure why my iPhone shows an option to translate website but all the destination languages to pick from (I have multiple languages installed), including English, are greyed out. iPhone does support translating from Chinese (Simplified or Traditional), and the button to translate website isn’t greyed out like it is for unsupported/unrecognized languages. Might be an iOS 27 bug, because it is working on other websites?

Since the link is in Chinese: MiMo Code is Xiaomi’s AI agentic coding harness.

“ MiMoCode is a terminal-native AI coding assistant. It can read and write code, run commands, manage Git, and use a persistent memory system to keep a deep understanding of your project across sessions while continuously improving itself.”

GitHub link (English): https://github.com/XiaomiMiMo/MiMo-Code

@dang might be better to link to the GitHub, and not for language reasons.

(Edit: for posterity, original URL as submitted was [0]).

[0]: https://mimo.xiaomi.com/mimocode

Different restrictions. ML gets treated differently from the rest.

My big question as an OSS dev distributing some precompiled binaries via npm for easy installation: does allowScripts also default to disabled when directly installing a package (globally or otherwise)?

As usual, another release post with no mention of what the damn thing is.

“A modern, self-hosting Object Pascal compiler built for the 2020s. Zero legacy, full ARC, and unified UTF-8.”

This is bullshit on par with the Chinese firewall, meant to effectively prevent the (entire!) western world from information by parties deemed persona non-grata. SSL certificates are supposed to be about security, not geopolitics.

I'm pretty sure a LE server hitting an Iranian or North Korean endpoint and validating a crypto challenge does not break any OFAC or EAR rules, and no money changes hands. And if a non-US entity wants to do it, the US would just sanction them. Microsoft and Mozilla are certainly not going to include a North Korean or Russian state CA in the root trusted certs (and if they did, the US government could just threaten them with sanctions, too).

Hard not to say "we warned you" about making self-signed certs completely unusable in favor of a very centralized approach.

I used Gemini exclusively via the API but downloaded the app last week for something. Even on max settings, it is ridiculously nerfed!

It's saying it's better than naively truncating the QAT release to 4 bits.

I'm not surprised Chen's patch was rejected; that's an extremely niche usecase not worth supporting. With my shell developer hat on, I agree with the closing "developers would likely welcome a native implementation that isn't (unlike the current implementation) hiding fork() and exec() under the covers".

How were you getting anything useful out of that? We found the (unquantized!) E2B model to be completely useless at even the simplest real-world classification tasks.

So what we want now is unsloth (or anyone) to release 4/6-bit quantized models of these releases?

Why is this written as a marketing spiel instead of just, you know, answering the question?

ADBC: https://arrow.apache.org/docs/format/ADBC.html

Seems like a columnar version of ODBC, for OLAP instead of OLTP.

Quite aside from the architectural changes, I suppose this is the answer to why Google had such a glaring hole in the (pretrained) Gemma4 model lineup between the Gemma4 4b and Gemma4 26b models!

A model that comfortably fits in 16GB of VRAM (allowing room for context) is a welcome upgrade.

I’m not sure I agree? For each of your examples there are algorithmic approaches and neural network approaches. Companies have certainly been loose and wild with how they market these, but there remain distinct approaches and implementations for each. Very generally speaking, the neural network based approaches (aka “generative AI”) perform better but with much worse degenerative cases and a higher baseline rate of unwanted side effects (that are normally not immediately visible but tend to cause issues down the line).

My bigger concern is that these neural network based solutions have taken the place of the former rather than supplemented them. Many tools no longer provide the algorithmic/kernel-based approach at all, and have marketed the “AI” (née ML) alternative as a strict superset/upgrade, despite its potential drawbacks.

(Interestingly while the inference-based implementations generally have higher latency (or infinitely worse, cloud and pay-as-you-go requirements), for some computationally difficult kernels the inference-based approach is actually faster!

Microsoft has been releasing LLMs for years.

Is this a Google-only proposal? Has Mozilla provided their thoughts on the matter?

Why would the default language-level union keyword implementation force boxing? Seems like a crazy decision if it could just also implement the HasValue/TryGet itself and avoid it?

This one keeps crashing the browser after it reaches 100%. Safari/iOS, iPhone 13. I was able to navigate and use a few of the other ones linked to from the comments, though. Curious.

Good write up! Relevant folklore Usenet post about the history of the calendar I am again obligated to share: https://neosmart.net/forums/threads/an-extended-history-of-t...