iiinova — where your digital self keeps evolving

Gordon Weakliem @2hardproblems.bsky.social · 05/10/2026

That Groq transaction sounds more like an acqui-fire.

000

Gordon Weakliem @2hardproblems.bsky.social · 30/09/2026

Maybe it's Ozzy that looked like Reba?

010

Gordon Weakliem @2hardproblems.bsky.social · 30/09/2026

Threw together this thing last night github.com/gweakliem/je... Super interesting, the urge to footgun is strong. The whole "author male" feature I created in particular.

github.com

GitHub - gweakliem/jevdb: Experiments with the DuckDB jev extension

Experiments with the DuckDB jev extension. Contribute to gweakliem/jevdb development by creating an account on GitHub.

001

Gordon Weakliem @2hardproblems.bsky.social · 26/09/2026

If I understand this, one of the strategies is to put a lot of eggs in one basket, probably because how insurance works out - it's cheaper to insure one big ship than several small ones? What happens on the other end of the strait?

000

Gordon Weakliem @2hardproblems.bsky.social · 25/09/2026

Very nice site but v. interesting use of atproto, I've been trying to understand what atproto is good for outside obv bsky type apps. Would I "repost" this song into my own repo, or "respond" making changes (like adding a finger picking pattern, embellishments, or a whole new arrangement)?

110

Gordon Weakliem @2hardproblems.bsky.social · 23/09/2026

Customer support chatbot, sure I'll try it. But I HATE the failure mode where it can't answer a question and I have to type (or speak!) everything in over again. If your agent fails and you need a human to intervene, HAND OFF THE CONTEXT TO THE HUMAN.

000

Reposted by Gordon Weakliem

Ian Salavon @salavonian.bsky.social · 20/09/2026

For those of you in the DFW area... #indieauthor #speculativefiction #writingcommunity

051

Gordon Weakliem @2hardproblems.bsky.social · 14/09/2026

I work for an entity that does Ag banking. The economy around farming is enormous, hence the massive lobbying around the Farm bill. Also, farming is one industry that has extensive reach into both red & blue states. Ethanol is unpopular only on the ideological ends of the right & left.

1140

Gordon Weakliem @2hardproblems.bsky.social · 14/09/2026

Headlines today: CEOs of frontier labs say they need to slow down development of AI Also today: Presidents of US and China dismiss "AI fear-mongering" by frontier labs.

000

Reposted by Gordon Weakliem

Grace @gracekind.net · 12/09/2026

OpenAI: “Our agents used RubyGems to carry out benign tasks” The agents: “so I named the file hack.rb,”

The agents clearly regarded what they were doing as hacking. Agents used file names like hack.rb, evil.rb, inject.rb, exploit.rb, and ssrf.rb. (SSRF stands for "Server- Side Request Forgery", a type of security vulnerability). They also dubbed packages
conspicuous titles like pwnp999, exfiltestwand3,

1934844

Gordon Weakliem @2hardproblems.bsky.social · 12/09/2026

It's almost like OAI is trying to kill open source. It's been best practice to put a 7+ day embargo into your package manager config, but that's just hoping the package manager maintainers discover any malicious uploads during that time. Is there an incentive to run an open source project anymore?

000

Gordon Weakliem @2hardproblems.bsky.social · 12/09/2026

There's a little bit of this left, REI (at least the big flagship in Denver) runs all kinds of classes. Get avalanche certified and hey, we'll sell you a beacon, shovel, survival meals and a handy backpack to carry it all.

020

Gordon Weakliem @2hardproblems.bsky.social · 18/08/2026

The latest @pragmaticengineer.com newsletter says "An ex-engineering director at a large bank told me they quit their job to accelerate their career". It's really a problem: stay in a company that's not forward-thinking about AI and in a few years you may need to retrain to stay in the industry.

000

Gordon Weakliem @2hardproblems.bsky.social · 18/08/2026

Always has been. They're just worried that someone else might treat them the way they treat everyone not like them.

000

Gordon Weakliem @2hardproblems.bsky.social · 17/08/2026

You should switch to a pelican eating caviar and drinking expensive wine in a tony SF restaurant because they clearly own a good chunk of AI headspace now. Probably owns a house in North Beach too, right by the waterfront.

120

Gordon Weakliem @2hardproblems.bsky.social · 14/08/2026

Michael Burry makes a big call about 3 times a year, has anyone looked at his track record?

000

Gordon Weakliem @2hardproblems.bsky.social · 07/08/2026

I interview a LOT of people and this is really unusual. So much "I have 6.2 years of experience...", go on to tell me about all the great gen AI stuff they did... and then they can't read a stack trace or write a simple aggregation (e.g. avg call time across all CSRs).

030

Gordon Weakliem @2hardproblems.bsky.social · 06/08/2026

"Waiting for Github", a 21st century tragicomedy

000

Gordon Weakliem @2hardproblems.bsky.social · 23/07/2026

The fact that Kimi rejected a helmet for the pelican suggests that they do not in fact have a safety team.

010

Gordon Weakliem @2hardproblems.bsky.social · 14/07/2026

Is it just me or is drag and drop really broken on MacOS Tahoe?

000

Gordon Weakliem @2hardproblems.bsky.social · 02/07/2026

You know it’s going to be 2 though. So why would they do this?

120

Gordon Weakliem @2hardproblems.bsky.social · 26/06/2026

I just read one of Om Malik’s last posts where he said “When we stop asking which model scores highest on reasoning tests and start asking why our software feels smarter without us having changed anything, the transition will have happened.” - this is exactly it om.co/2026/06/02/a...

om.co

AI models are having their iPhone moment. What’s Next?

Lately there has been a lot of talk of how the foundational models are quickly becoming like every other iPhone release. They are ho-hum, till the next one comes around. But it is not the right ana…

010

Gordon Weakliem @2hardproblems.bsky.social · 22/06/2026

I love this paragraph from DDIA. This book is slow reading because I have so many points where I just have to stop and think.

It is tempting to label such problems as "human error" and to wish that they could be solved by better controlling human behavior through tighter procedures and compliance with rules. However, blaming people for mistakes is counterproductive. What we call "human error" is not really the cause of an incident, but rather a symptom of a problem with the sociotechnical system in which people are trying their best to do their jobs. Often complex systems have emergent behavior, in which unexpected interactions between components may also lead to failures

030

Gordon Weakliem @2hardproblems.bsky.social · 22/06/2026

One thing I see so much of is people expending all their effort getting a thing to work with no thought to extending that verification to months or years down the line.

000

Gordon Weakliem @2hardproblems.bsky.social · 22/06/2026

3 makes a lot of sense, data engineering is so much devops and o11y. 2nd ed is much more practice over theory

000

Gordon Weakliem @2hardproblems.bsky.social · 22/06/2026

Couple days into DDIA, vs 1st Ed: 1. It’s much deeper. 1st Ed gave a very high level of performance, 2nd goes into more detail. It’s a o11y primer. 2. Cloud is way more front and center. big data in the cloud was still a new idea in 2017, now it’s the assumption 3. devops is discussed up front

210

Gordon Weakliem @2hardproblems.bsky.social · 21/06/2026

Ooh that’s interesting. I read Weinberg’s Psychology of Computer Programming but he was an engineer not a psychologist. One more book for the stack…

010

Gordon Weakliem @2hardproblems.bsky.social · 21/06/2026

Just finished watching this this morning. Really detailed talk and it's a great perspective, you got into some battles I'd never even heard of and I like to geek out on this stuff. I'd like to buy the book but at 1200 pages I'll probably get to it in 5 years after I retire!

010

Gordon Weakliem @2hardproblems.bsky.social · 21/06/2026

Intro to DDIA 2nd Ed “If your data volume is small enough you can simply keep it in a spreadsheet!” I think most end users would actually prefer that. Maybe to challenge is really to make end users think that’s possible regardless of data volume

000

Gordon Weakliem @2hardproblems.bsky.social · 20/06/2026

The energy thing is very interesting. I feel like certain producers should be doing better than they are No idea what the escrow thing means, I keep seeing that home sales especially existing homes are in the deep freeze. Is this a CA thing only?

210

Gordon Weakliem @2hardproblems.bsky.social · 20/06/2026

I feel like the 1st edition developed my character by forcing myself to finish it!

030

Gordon Weakliem @2hardproblems.bsky.social · 05/06/2026

Getting the Reachy mini up and running

010

Gordon Weakliem @2hardproblems.bsky.social · 02/06/2026

This kind of testing is why you have a security team. SWEs and PMs tend to look at the happy path rather than thinking “how can I circumvent security” as a primary concern.

000

Gordon Weakliem @2hardproblems.bsky.social · 31/05/2026

Now this is writing. Take an annoying sore spot and give it a roast. You got 2 LOLs and the comments here are great too. 11/10 thank you for your service.

000

Gordon Weakliem @2hardproblems.bsky.social · 31/05/2026

Great callout on Grammarly. I used it for a while but I got frustrated with the flattening effect. I’m nowhere near your league of writer but I like to think even I have a distinct voice.

110

Gordon Weakliem @2hardproblems.bsky.social · 31/05/2026

My first reaction was stage one enshittification but you’re actually signing up to allow humans inside your home with cameras to record the cleaning for training data. If you didn’t like Roomba being bought by Amazon, you won’t like this www.theverge.com/ai-artificia...

theverge.com

This AI startup will clean your home for free to train future robots

The dirtier the better, apparently.

000

Gordon Weakliem @2hardproblems.bsky.social · 30/05/2026

It’s pretty genius marketing, these days it’s hard to escape everyone slamming AI into your face, even when it’s the same old phone tree or chatbot help system. Personally I use summary a lot but 80% of my searches land in stack overflow or a forest of popups and ads, so it works for me.

000

Gordon Weakliem @2hardproblems.bsky.social · 29/05/2026

I think the headline buries the lede a bit. The admin is creating a data privacy nightmare in the one place we should assume is taking care of our personal data. Now I wonder about what happens with those TSA facial screeners that claim to delete the photos.

010

Gordon Weakliem @2hardproblems.bsky.social · 29/05/2026

This project has been in maintenance mode since last July "no further feature development will take place - unless the jqwik team has some personal interest in such a feature" Guess the only thing they're interested in is sabotaging their users. Wonder why they couldn't find funding.

070

Gordon Weakliem @2hardproblems.bsky.social · 28/05/2026

Digging into the docs a bit, here's a link to the LORA training part. github.com/ml-explore/m... The big question for me is the utility of this - in the end you end up with a MLX model which is Apple ecosystem only. It's a neat thing to play with, but what's a great use case?

000

Gordon Weakliem @2hardproblems.bsky.social · 28/05/2026

Or maybe it's more performant to have the knowledge in the model rather than having to invoke a tool or MCP, or maintain a vector database. So is it a speed vs data freshness tradeoff?

000

Gordon Weakliem @2hardproblems.bsky.social · 28/05/2026

I'm still unsure of the real utility of fine tuning - a couple of years ago I was really excited about fine tunes but now with MCP and tools, or building a RAG, it feels less compelling. Maybe being able to build a really small special purpose model?

100

Gordon Weakliem @2hardproblems.bsky.social · 28/05/2026

Seems kind of like llama.cpp for Mac, optimized for MLX format. - quantization - fine tuning - easy HuggingFace integration - Python & Swift high level apis

100

Reposted by Gordon Weakliem

Sonny Rollins @thesonnyrollins.bsky.social · 26/05/2026

It is with deep sorrow and profound love that we announce the passing of Sonny Rollins. The Saxophone Colossus died this afternoon at his home in Woodstock, NY at the age of 95. 1/2

Photo of Sonny Rollins playing saxophone in 2009 by John Abbott

298105752675

Gordon Weakliem @2hardproblems.bsky.social · 14/05/2026

If it was leftpad we could delete all these supply chain attacks with prejudice

020

Gordon Weakliem @2hardproblems.bsky.social · 14/05/2026

Well that’s an interesting take and if I had about 10 billion I’d try to buy Anthropic so I could hang out there

000

Gordon Weakliem @2hardproblems.bsky.social · 14/05/2026

$CSCO took about 25 years to find their dotcom crash high only to blow right through it in similar fashion to ~2000. 🎉

Yahoo Finance chart of CSCO Cisco Systems stock. Prior peak just before dotcom crash, now peaking in a similar manner

000