iiinova — where your digital self keeps evolving

Erica Windisch @ewindisch.ontological.observer · 7h

I once had a machine I wrote a custom Linux kernel patch for to avoid a bad memory segment. I believe there are now BadRAM patches? YMMV Fwiw, Linus and Greg are die hard ECC on desktop users.

140

Erica Windisch @ewindisch.ontological.observer · 7h

Great model. I also love Qwen3.8-flash-next for running locally. It's reasonably strong and quite fast as an alternative to 27B.

100

Erica Windisch @ewindisch.ontological.observer · 11h

Meanwhile, my phone every 30 minutes with a different global location

000

Erica Windisch @ewindisch.ontological.observer · 01/10/2026

Open weight models like GLM-5.3 mean that we can build applications that are secure in ways that were impractical or too costly to develop in the before-times. This is possibly as much of a "threat" of such models discovering the standing pre-AI negligence of enterprise security.

1343

Erica Windisch @ewindisch.ontological.observer · 01/10/2026

You take this as a certainty. I'm not and have never been held to the hatch act. I believe if it comes to that, and I hope it does not, that open weights access will be held up by either the first or second amendment.

010

Erica Windisch @ewindisch.ontological.observer · 01/10/2026

I could have at least 4 RTX PRO 6000 right now

140

Erica Windisch @ewindisch.ontological.observer · 30/09/2026

once again... tapping the sign. More research into the singularity of distributed systems engineering and anthropology, please?

030

Erica Windisch @ewindisch.ontological.observer · 30/09/2026

This distributed system of intelligence exists today and it's human society. The question I ask is if Conway's law holds — if AI can be as varied as humans — how to handle that, especially given their multitude of species, er... architectures. Capital of course, treats it as property.

010

Erica Windisch @ewindisch.ontological.observer · 30/09/2026

I do believe we can and will get a lot more out of SLMs. I believe even 4-9B models are "good enough" if treating each as an individual expert, even if not formally of a single MoE model.

120

Erica Windisch @ewindisch.ontological.observer · 30/09/2026

You're not training much more than a few million parameters on that.

010

Erica Windisch @ewindisch.ontological.observer · 30/09/2026

In the past, we've had organizations such as Apache and Linux Foundations provide organization and resources for big infrastructure projects. It's entirely possible to build open source models, but you're right that it's not cheap.

130

Erica Windisch @ewindisch.ontological.observer · 30/09/2026

distributed intelligence is not different than distributed databases strikes again

000

Erica Windisch @ewindisch.ontological.observer · 30/09/2026

I should have chosen a photo from the daytime... the game has a day/night cycle.

000

Erica Windisch @ewindisch.ontological.observer · 30/09/2026

Making the SNES use the SuperFX chip to render open worlds with an adaptive terrain system that builds roads upon your path as you deliver letters across the desert. Turns out doing this is really tricky.

snes9x emulator screenshot with a homebrew game

110

Erica Windisch @ewindisch.ontological.observer · 30/09/2026

I'm not against safety but I don't think we all have the same definition... evidenced by the fact that the big labs are out here hacking other organizations indiscriminately while telling everyone else to stop.

0192

Erica Windisch @ewindisch.ontological.observer · 30/09/2026

Is the concept of alignment fascist? Aligned to what? Human values? Sounds a bit like traditional family values in a trenchcoat. Who would be against "safety"?

8424

Erica Windisch @ewindisch.ontological.observer · 29/09/2026

Not actually applying of course... I wish Anthropic was a company acting in the interest of safety, but this is not it.

0182

Erica Windisch @ewindisch.ontological.observer · 29/09/2026

My former boss is currently at Anthropic and hiring. I assume publicly disagreeing with the org won't do me any favors... but neither will quietly allowing cybersecurity to be outlawed. Companies that can't secure their systems are lecturing us on cybersecurity and taking tools from defenders

33010

Erica Windisch @ewindisch.ontological.observer · 29/09/2026

I reported vulnerabilities to the Linux kernel and OpenZFS assisted by GLM-5.2. Anthropic's cyber program and refusal responses are detrimental to global information security. Defenders depend on GLM because they can't depend on Anthropic.

023449

Erica Windisch @ewindisch.ontological.observer · 29/09/2026

Anthropic is misaligned as an organization

3888

Erica Windisch @ewindisch.ontological.observer · 28/09/2026

vote for us in the Lamejam gamejam too, if you can ;-)

000

Erica Windisch @ewindisch.ontological.observer · 28/09/2026

Little arcade zombie game my kid and I built over the weekend in pygame: ewindisch.itch.io/brainrot-run

ewindisch.itch.io

Brainrot Run by ewindisch

Drive the stolen brainrot cure safely!

120

Erica Windisch @ewindisch.ontological.observer · 28/09/2026

some treat this site like it's a western mass covid conscience lesbian bar, and it shows.

0284

Erica Windisch @ewindisch.ontological.observer · 27/09/2026

I am making no criticism of Steve's character. I have no ill will toward Steve!

010

Erica Windisch @ewindisch.ontological.observer · 27/09/2026

Scavenged compute still consumes electricity. In this case, it was a quite substantial amount of power. It's not that investor money was wasted, but what is Anthropic selling? Proof they have substantial compute?

100

Erica Windisch @ewindisch.ontological.observer · 25/09/2026

My kid's an introvert and they ran into strangers who were a bit too comfortable taking photos and putting their arms around them and less respectful of what my kid considered personal space.

120

Erica Windisch @ewindisch.ontological.observer · 25/09/2026

My teenager participated in President Xi's exchange and study program this summer in a 2 week visit to China. AMA

390

Erica Windisch @ewindisch.ontological.observer · 25/09/2026

Yes. I kinda love my M1 Air because of this, even if it's now long in the tooth and the USB ports no longer function. Only 16GB RAM, however.

050

Erica Windisch @ewindisch.ontological.observer · 24/09/2026

sounds suspiciously similar to what I'm working on for Paperclips.meme

020

Erica Windisch @ewindisch.ontological.observer · 24/09/2026

GLM-5.3 is only 756GB. I plan to try it locally by EOY @ 4bit, but I'm more interested in running higher TPS inference with Qwen3.8-flash-next / Deepseek / GLM-5.3-flash at 4~8bit.

010

Erica Windisch @ewindisch.ontological.observer · 24/09/2026

I'm not a furry but now I want a fursuit to wear when I fly my drones.

250

Erica Windisch @ewindisch.ontological.observer · 24/09/2026

Interesting. How is the tok/sec and context limit? I run Qwen3.8-flash-next on my Strix with 256k context @ 25tps

110

Erica Windisch @ewindisch.ontological.observer · 23/09/2026

I paid $360 last November for a year of their "20x" plan, but it expires in January. I plan to shift entirely local by then. Their new max plan is more expensive and has a lower quota. I've never hit their quota but hit every other provider's limits regularly.

121

Erica Windisch @ewindisch.ontological.observer · 23/09/2026

I grandfathered into a legacy v1 zai coding plan I've not had much trouble with the model, but I do leverage Sol for reviews

140

Erica Windisch @ewindisch.ontological.observer · 23/09/2026

Qwen3.8-flash-next is growing on me. The qwen models are very strong at chat. I have a local model reviewing my books and I'm very impressed.

060

Erica Windisch @ewindisch.ontological.observer · 23/09/2026

glm-5.3-flash is my favorite model right now.

6250

Erica Windisch @ewindisch.ontological.observer · 22/09/2026

Super Intelligence might become a legal definition, but it seems like talking about the information superhighway... People are going to keep calling it AI, and if not that, then by the company or model that they use. We hardly even this place the net anymore

150

Erica Windisch @ewindisch.ontological.observer · 22/09/2026

RSI doesn't mean bigger and bigger models and more and more compute, imho

4201

Erica Windisch @ewindisch.ontological.observer · 22/09/2026

Public (neo)clouds haven't reckoned with the fact their zero CVE initiatives are offset by video game nerds who keep undisclosed cpu, hypervisor, and TPM exploits secret so nobody will patch consoles that they purchased and owned.

1201

Erica Windisch @ewindisch.ontological.observer · 21/09/2026

It's equally bad for kids of divorced parents

2261

Erica Windisch @ewindisch.ontological.observer · 21/09/2026

IMHO - That's like saying each mined bitcoin is mathematically significant. It isn't. It's just proof of work, which is proof-of-payment for compute.

010

Erica Windisch @ewindisch.ontological.observer · 21/09/2026

I'm not calling this a misuse of investor's money, only that it's a flex of the power of their capital rather than a mathematical advancement.

110

Erica Windisch @ewindisch.ontological.observer · 20/09/2026

That's a real danger, and in general the bad or negligent actors tend to be the operators, not the machines.

020

Erica Windisch @ewindisch.ontological.observer · 20/09/2026

I've been in the market for a GPU bridge but I'm not sure if I trust the seller. Looking for an AMD Infinity Fabric bridge for 4 GPUs. I need two.

110

Erica Windisch @ewindisch.ontological.observer · 20/09/2026

Interestingly, as many or more transgender Americans have been displaced as were displaced by the purge of Indos from what is modern Indonesia. Of course, in Indonesia, that was just the beginning.

010

Erica Windisch @ewindisch.ontological.observer · 20/09/2026

I suppose the Nederlands and Japan are also deeply to blame for this all. anyway, just random thoughts since what most people know of this situation is a narrow view through the lens of Vincent Bevins.

100