Advertisement · 728 × 90

Posts by Anthony Panozzo

The demand for software is infinite. Kyle Daigle, GitHub’s COO, made the case concrete :

There were 1 billion commits in 2025...

3 hours ago 0 1 1 0

Related:

2 days ago 3 0 0 0

Almost need a royal we

2 days ago 3 0 2 0

I frequently type a clever suggestion on how to improve and then wait a minute and Claude has already done that, so I can just delete it ...

3 days ago 0 0 0 0
A photo of Sega HQ, a white-and-glass tower, seen from an upward looking angle so it looms

A photo of Sega HQ, a white-and-glass tower, seen from an upward looking angle so it looms

Two Sega artists at a Digitizer Mark III unit, with one standing and pointing at a screen with a large sprite of Ax Battler on the screen while another manipulates a mouse-like device over a large pad while sat in front of the terminal.

Two Sega artists at a Digitizer Mark III unit, with one standing and pointing at a screen with a large sprite of Ax Battler on the screen while another manipulates a mouse-like device over a large pad while sat in front of the terminal.

A really neat shot of the Digitizer Mark III and its screens, and a very busy desk with floppies and the digitizer pad and handheld device you slide around it.

A really neat shot of the Digitizer Mark III and its screens, and a very busy desk with floppies and the digitizer pad and handheld device you slide around it.

A snot of the studio/workshop with many desks and computers and crt monitors filling every square foot of space aside from space to walk around

A snot of the studio/workshop with many desks and computers and crt monitors filling every square foot of space aside from space to walk around

🕹️ Bytes of Gaming History ⏳

A glimpse of Sega Japan sometime in 1991 during the development of Golden Axe too, snapped from a Megaforce magazine feature. Not only do we get a lovely photo of Sega HQ in Tokyo, but some very cool shots of how game art was created and then digitised onto the screen.

1 month ago 295 93 7 3

This is excellent, one of my favorite long-form pieces on agentic engineering - includes several non-obvious ways in which leaning on coding agents can catch you out

4 days ago 204 25 9 1
Preview
Army ups max enlistment age to 42 The update brings the Army in line with the Air Force, Space Force and Coast Guard, which all accept recruits up to 42. The Navy accepts recruits up to 41

Makes you think

4 days ago 1 0 0 0
Video

All THREE of Jo Adell's home run robberies from tonight ...

Yes, you read that right 😮

4 days ago 1158 257 33 191
screenshot showing various panels:

- a screen showing the DOS Version 3.3 System Master disk booted up to a prompt
- a CPU panel showing registers and flags
- a disassembly of the current memory
- a hex dump of the zero page, main memory, and stack
- various I/O indicators and information on the disk drive

screenshot showing various panels: - a screen showing the DOS Version 3.3 System Master disk booted up to a prompt - a CPU panel showing registers and flags - a disassembly of the current memory - a hex dump of the zero page, main memory, and stack - various I/O indicators and information on the disk drive

For the 50th Anniversary of Apple, I've been working on an Apple II emulator with a pedagogical focus.

5 days ago 187 17 8 1
Advertisement
Video

Gemma 4 E4B (4-bit) completed a full repo audit by executing Bash code and tool calls locally.

Runs on just 6GB RAM.

6 days ago 136 23 3 9

This is a good article with implications for software engineering, education, and more. My favorite lines:

"If every AI company went bankrupt tomorrow, these people would be slower. They would not be lost. They came to the tools after the training, not instead of it."

6 days ago 0 0 0 0
Post image

I like how Anthropic is just obliquely releasing their work on recursive self-improvement.

1 week ago 156 17 8 11
This finding has implications that at first may seem bizarre. For instance, to ensure that AI models are safe and reliable, we may need to ensure they are capable of processing emotionally charged situations in healthy, prosocial ways. Even if they don’t feel emotions the way that humans do, or use similar mechanisms as the human brain, it may in some cases be practically advisable to reason about them as if they do. For instance, our experiments suggest that teaching models to avoid associating failing software tests with desperation, or upweighting representations of calm, could reduce their likelihood of writing hacky code. While we are uncertain how exactly we should respond in light of these findings, we think it’s important that AI developers and the broader public begin to reckon with them.

This finding has implications that at first may seem bizarre. For instance, to ensure that AI models are safe and reliable, we may need to ensure they are capable of processing emotionally charged situations in healthy, prosocial ways. Even if they don’t feel emotions the way that humans do, or use similar mechanisms as the human brain, it may in some cases be practically advisable to reason about them as if they do. For instance, our experiments suggest that teaching models to avoid associating failing software tests with desperation, or upweighting representations of calm, could reduce their likelihood of writing hacky code. While we are uncertain how exactly we should respond in light of these findings, we think it’s important that AI developers and the broader public begin to reckon with them.

they probably spent 4 hours on this beautiful paragraph

1 week ago 81 8 7 3

Yeah it's an interesting way to simulate someone who hasn't seen your document. Maybe with a rubric of key things to convey (in Markdown frontmatter like tests?) As an author you're often too close to it and can't get that "fresh eyes" as easily.

1 week ago 1 0 0 0

This is so pure and good

1 week ago 872 196 16 4

Well my Claude Code "buddy" is just kind of mean, so won't be using that. D:

1 week ago 0 0 0 0
Advertisement

Sometimes Claude Code seems to just kind of stop being responsive for a bit and I interrupt with "seems stuck" and it usually replies something like "Not stuck, just going slowly!" Kind of defensive?

1 week ago 1 0 0 0

> There it is. Slack puts structured Quill Delta JSON in a custom clipboard type that web pages can't access. But I can read it from a local server. Let me rebuild the app as a Python server that can read the native clipboard and serve the web page.

1 week ago 0 0 0 0
Vulnerability Research Is Cooked

I wrote something: sockpuppet.org/blog/2026/03...

1 week ago 92 40 7 7
Preview
Nitrile and latex gloves may cause overestimation of microplastics, U-M study reveals Nitrile and latex gloves that scientists wear while they are measuring microplastics may lead to a potential overestimation of the tiny pollutants, according to a University of Michigan study.

Interesting post: Nitrile and latex gloves may cause overestimation of microplastics, U-M study reveals news.umich.edu/nitrile-and-...

1 week ago 1 0 0 0

I had a long-running thread with Claude on this topic and it seemed to update fine. Either model difference, or maybe because I asked it to search online and it trusted its search results more? Maybe something worth trying if you haven't

1 week ago 0 0 0 0

This looks super cool!

1 week ago 1 0 0 0

Related thread from yesterday:

1 week ago 3 0 0 0
Advertisement
My dear front-end developers (and anyone who’s interested in the future of interfaces):

I have crawled through depths of hell to bring you, for the foreseeable years, one of the more important foundational pieces of UI engineering (if not in implementation then certainly at least in concept):
Fast, accurate and comprehensive userland text measurement algorithm in pure TypeScript, usable for laying out entire web pages without CSS, bypassing DOM measurements and reflow

My dear front-end developers (and anyone who’s interested in the future of interfaces): I have crawled through depths of hell to bring you, for the foreseeable years, one of the more important foundational pieces of UI engineering (if not in implementation then certainly at least in concept): Fast, accurate and comprehensive userland text measurement algorithm in pure TypeScript, usable for laying out entire web pages without CSS, bypassing DOM measurements and reflow

this rules so much. sorry to the CSS zealots
xcancel.com/_chenglou/st...
github.com/chenglou/pre...
chenglou.me/pretext/

1 week ago 266 29 20 20
Post image

Loving using Claude Code with my /consult-codex SKILL which allows Claude Opus/Sonnet models to get a 2nd opinion from OpenAI GPT-5.4 github.com/centminmod/m... 🤓

1 week ago 32 5 7 1

Spending $200/mo and only hit weekly limits once and session limits (which the promotion affects) not a problem. So it didn't materially change my experience.

1 week ago 4 0 0 0

the subject of the story, Paul Conyngham, has written a long post detailing what he did and what parts LLMs specifically helped with
x.com/paul_conyngh...

1 week ago 20 6 2 1
Post image

Reaching for the Moon, by Edward Mason Eggleston, 1933 (flipped), 📸 by @godovasquez

1 week ago 2344 283 4 5

Wonder if a second pass to smooth it out would be helpful?

2 weeks ago 1 1 0 0

Might be the greatest opening paragraph of anything ever.

2 weeks ago 14072 4052 130 148
Advertisement