Gidi
@gidim.bsky.social
200 followers 820 following 250 posts
Enough. חלאס. خاس Nature advocate striving to halt biodiversity loss & tackle climate change 🌿. Experienced Head of Engineering and technical founder in the NatureTech space. Thoughts on Engineering at gidi.io. שלום عَلَيْكُمْ‎ ✌️
Posts Media Videos Starter Packs
gidim.bsky.social
As someone who worked in Ad Tech I can tell you that with high likelihood Ads are the culprit.
gidim.bsky.social
Sad to say I'm seeing as much racism and antisemitism on bluesky as I was seeing on Twitter before I left it.

Might be time to just switch it all off.
gidim.bsky.social
Pitkah Tovah.
May we have less to reflect on next year.
gidim.bsky.social
I have a fear, which might be totally unfounded but I guess we'll find out, that folks are going "AI first" when trying to solve problems before they even bother trying to figure it out.

If that's the case, I worry engineers are going to lose their problem solving skills very quickly.
gidim.bsky.social
Haha yeah, I guess, for boomers and gen X it's a smaller proportion of the generation so less representative....

But yeah, bad generalisation.
gidim.bsky.social
I think we're the only generation that had to figure out how things worked to get basic things done on computers....

Though they might be a bias on my end.
gidim.bsky.social
I was hiking in fjordland last year and saw a flock of parrots above... Expecting a mischievous Kea to decend I felt for my car keys when a unusual screech came out above - it was no Kea, it was a Kaka. ☺️

My first time spotting them out in the wild.
gidim.bsky.social
Schemaless databases lack a schema in the same way dynamic languages lack a type system.
That is to say, they don't, it's just abstracted away.

As a result, you still have migrations, and they are often harder.

I always start with postgres unless I *know* it makes the wrong tradeoffs for my domain
gidim.bsky.social
Same, I've always used the EM dash in my writing but—if I'm being honest—I use it way more now that folks are using it to misclassify content as LLM generated.
gidim.bsky.social
Haha, I wish you health to use it.
gidim.bsky.social
This definitely falls into the category of "notebooks I'll buy but never use because I'm waiting for a worthy enough idea".

I already have a half dozen in this category. 😅
gidim.bsky.social
I've long felt that testing is a dying art, but LLMs feel like the a stab to the heart of this art.

Its unfortunate, because I actually find LLMs pretty good at generating the green implementation for my red failing TDD tests. 🤷

But most engineers don't TDD so it's just generating noise
gidim.bsky.social
I would guess aprox 95% of the tests that candidates have submitted to our take home exercise have been generated by LLMs.

The telltale signs are pretty obvious.

Its pointless to test that your implementation does what it does. You should be testing that implements the business logic.
gidim.bsky.social
I glare at the terminal in disgust. 😑

The tests are wrong. 🙄

I haven't implemented the code yet. 🤨
gidim.bsky.social
Haha, not my description but theirs 😉.
gidim.bsky.social
Not exactly.

To better gauge hype versus reality, its a visualization showing the distance between the proportion of respondents who want to use a technology (“🔵desired”) and the proportion of users that have used the same technology in the past year and want to continue using it (“🔴admired”).
gidim.bsky.social
When people tell you who they are - believe them.

Reform are climate change deniers and they are a real threat to our future existence.

If you know anyone who is considering voting for them, make it your personal project to help them understand what they are voting for.
roberthutton.co.uk
Christopher Monckton says that under a Farage government, "there is a sporting chance that you are now looking at the Secretary of State not for but Against Climate Change Nonsense.”
gidim.bsky.social
Agreed, new grads are high ROI - but many employers hire grads for the low paycheck and then fail to invest in them, resulting in little to no ROI.

I don't want to further reduce the pool of jobs for new grads - but folks should only expect to see a return if they're willing to invest.
gidim.bsky.social
This gives me two thoughts:

1. They're probably midway through a fundraise and the CEO is pushing because he sold investors a story about them being an AI company

2. I should go back and ensure my ancient coinbase account is full wiped because they are definitely going to have an infosec incident
gidim.bsky.social
Yes, obviously - that was actually not the problem in our case.

Our concern was whether it's *actually* doing what stakeholders thought it was doing (surprise, surprise, it wasn't) and whether we can fix that without breaking a bunch along the way (it's a complex domain and the code was terrible).
gidim.bsky.social
We're now at a place where most the code is still a black box that we don't touch, but the area where we had to make changes is refactored, testable and we now know what it's doing.

Its changed how I think about working with legacy code. A silver lining on the otherwise shitstorm that are LLMs. 😅
gidim.bsky.social
We used an LLM to backfill tests and then reviewed the tests to confirm we understood what each one was testing, and translated that into a "tested requirement".

Its still legacy. It's still shit. But at least we can refactor with a *semblance* of confidence we haven't broken anything.
gidim.bsky.social
Interestingly, I've found this really useful when taking over some terrible legacy code recently.

It had no tests, no clear docs, but stakeholders were adamant it's correct.
My team wasn't comfortable making changes as they had no way of knowing if they introduced a regression.
Reposted by Gidi
jasongorman.bsky.social
Generating tests based on your implementation is like setting exam questions based on your answers.