Bunheads, Game of Thrones, Wire, Deadwood, Mad Men. Archer. NBA.
The Danger Zone
- Tweets 108,258
- Following 4,924
- Followers 2,458
- Likes 190,109
Pinned Tweet
Rhaegar tweeted valiantly, Rhaegar tweeted nobly, Rhaegar tweeted honorably. And Rhaegar's battery died.
8
73
166
This tweet pairs almost perfectly with the G Orsoz RT (above, below).
Adj in timeline space.
It is the tweet about LLMs when you don't know the domain.
an important lesson i learned in grad school: just because someone sounds smart and speaks with confidence doesn’t mean they know what they’re talking about
1
25
House of Black&White retweeted
This messes with my mind ALL THE TIME
When I do stuff with AI I am an *expert* in (eg coding on my project I know well, or doing research in an area I am very familiar with) I spot problems/errors/the model BS'ing me immediately. No big deal, I have it correct or start again
BUT when I'm not an expert in an area, I find that I trust it, and then later learn it does stupid stuff. Like a case when I asked 3x AI models to give feedback on a legal case (that I also paid a lawyer for, but lawyer was slow). All 3x models confidently told me X, lawyer then told me Y, and lawyer was right
Claude Opus 5.5 is great, and most of the time, the solution is similar to how I'd implement it myself.
However, every now and then, I find terrible design decisions that would fail in production, and that's when I realize how important software engineering is, and will be, years from now.
133
73
1,492
80,056
House of Black&White retweeted
Every single Eng leader out there is iterating on some kind of PR autoapprover, coding automations, AI cost optimization, and model switching/eval stuff.
Just a million people doing the same thing.
37
22
899
55,260
House of Black&White retweeted
The games we've long wanted, which didn't exist because they'd cost several times more to make than they'd earn back, will soon become possible. As will the games developers wanted to build, but couldn't with the resources available. It'll be a new golden age of gaming.
2,122
1,272
17,206
5,488,796
House of Black&White retweeted
Unfinished coding side projects are now way easier to complete than unfinished blog posts.
I’m perfectly OK with AI writing my code, even the docs.
But I flinch at the idea of AI writing my blog posts, which feels like an egregious violation of a fundamental ethnical contract.
25
31
600
11,642
House of Black&White retweeted
Given the acceleration Anthropic is reporting in coding, it is amazing that you still cannot change your email address in Claude and the official advice is that if you need to change your email the only way to do that is delete your account.
209
82
2,214
93,189
House of Black&White retweeted
In this new age of LLMs, it feels very much like:
1. we are seeing a dramatic increase in the amount of new code being generated
2. we are not getting an even noticeable increase in the quality of software (from an end user point of view)
75
63
1,712
27,753
House of Black&White retweeted
A few of us have been predicting saturation of adversarial contact points for a while, but this is first set of good graphs I’ve seen of it.
Imagine it will take owning agencies a very substantial amount of time to realize what is happening.
To give an indication of the problem here's a chart of 12 government agencies around the world and numbers of applications / appeals / complaints etc...
The dotted grey line is when ChatGPT was invented.
24
53
696
36,900
House of Black&White retweeted
One of the biggest mistakes developers are making in the 2026 job market is confusing a finished project with proof of engineering ability.
Y'all need to understand something:
Nobody is hiring you because your portfolio has a cool application. They ARE hiring you because of the engineering conversations that application allows you to have.
Think about two candidates walking into an interview with the exact same AI-powered project.
- Candidate #1 explains what the application does, which AI tools they used, and shows off the features.
- Candidate #2 explains WHY they designed it the way they did. Where they used deterministic logic instead of AI. How they handle unreliable model outputs. What happens when the system fails. What tradeoffs they made and WHY.
Same exact project. Two completely different interviews.
One demonstrated that they could produce software.
The other demonstrated that they understood how to engineer it.
And this is why I keep pushing developers to understand the decisions they're generating with AI.
I don't really care if AI wrote 100% of your code. But if you can't explain the architectural decisions, defend your tradeoffs, or identify where things could go wrong, you've missed one of the biggest opportunities that project was supposed to give you.
AI lowered the barrier to building software. It did NOT lower the bar for understanding it and making the right decisions.
We got into this on the latest episode of The Programming Podcast because I think developers need to fundamentally rethink what they're trying to accomplish with portfolio projects.
#code #coding #programming #softwareengineering
17
36
210
13,103
It's me. I'm the bottleneck.
Anthropic’s report is basically saying humans are the bottleneck.
“Amid this model-driven acceleration, human attention is a constant bottleneck. Claude sessions amount to around 80% idly waiting for human input. As one engineer put it, people are “bottlenecking the agents.”
1
73
House of Black&White retweeted
NBA twitter: “These are all gather steps on the dribble.”
6
21
136
1,988
This is a question.
Different automation rates from diff user/inst profiles gonna cause friction.
A consequential choice will be made soon by government and companies as they are increasingly flooded with agentic AI inquiries that overcome the frictions that gated access (see the graphs - it has already started!)
Two options: do you make services worse or better as a result?
59
House of Black&White retweeted
toddlers are novelty search algorithms filling in an archive whose behavior descriptor is unimaginable and whose quality is a proxy for danger for things i never once considered before
1
5
459
This is so hard tho.
Working in large established codebases is one of the hardest things to learn as a software engineer. You can’t practice it beforehand (no, open source does not give you the same experience).
ALT Working in large established codebases is one of the hardest things to learn as a software engineer. You can’t practice it beforehand (no, open source does not give you the same experience). Personal projects can never teach you how to do it, because they’re necessarily small and from-scratch. For the record, when I say “large established codebases”, I mean: Single-digit million lines of code (~5M, let’s say) Somewhere between 100 and 1000 engineers working on the same codebase The first working version of the codebase is at least ten years old I’ve now spent a decade working in these codebases. Here’s what I wish I’d known at the start. The cardinal mistake is inconsistency
ALT There’s one mistake I see more often than anything else, and it’s absolutely deadly: ignoring the rest of the codebase and just implementing your feature in the most sensible way. In other words, limiting your touch points with the existing codebase in order to keep your nice clean code uncontaminated by legacy junk. For engineers that have mainly worked on small codebases, this is very hard to resist. But you must resist it! In fact, you must sink as deeply into the legacy codebase as possible, in order to maintain consistency. Why is consistency so important in large codebases? Because it protects you from nasty surprises, it slows down the codebase’s progression into a mess, and it allows you to take advantage of future improvements.
1
1
97
Pro sports, the coin flip should be replaced with rock paper scissors by team captains.
Back to back so they can't peek. Done.
42
What is it about these types of charts that seem so clear but in reality contain zero information?
The ghost of the data scientist in me couldn’t stand it. I asked Claude to rework itself, takes about 15 min
We keep improving performance across Claude, and you may have noticed it starts responding sooner!
Claude Code now gzips the conversation it sends on every turn, so it reaches the model faster. Cloud sessions also start up as soon as you start typing.
13
8
121
14,458