Shipped and scaled https://nitter.gq/t.co/mxwLCd1EOb and https://nitter.gq/t.co/QzKfNQlnMD. Still hacking, but a lot more surfing and skiing. Proud pops. Kooky.
Berkeley
Joined May 2009
- Tweets2.9K
- Following573
- Followers605
- Likes6.3K
Ooooh my immediate reaction to the new @github PR page layout is that it's MUCH better!
51
I know i'm not the only one saying this, but damn is Opus 5.5 on medium so much better at writing code than on a higher reasoning level. And much much better than Fable 5.1.
5
175
August Flanagan retweeted
My three suggestions for setting your agent-based software flow on fire:
1. Remove all skills, AGENTS.md, and CLAUDE.md. At minimum, ask Opus to remove everything not vital.
2. Use Claude desktop app. Sorry, yes.
3. Use Opus 5.5 Medium. It's significantly better than others.
August Flanagan retweeted
No matter which party controls Congress, spending goes up. The uniparty is in the driver’s seat, and the bills keep getting bigger.
My Six Penny Plan would cut six cents off every projected dollar that the federal government spends, and balance the budget in 5 years. It’s not complicated. Washington just needs the will to do it.
A 15 minute outage can feel like 15 hours. We just had our first big outage in a long time @monitorio. We're shipping at an incredible speed, and have invested heavily in automated testing and CI, and yet there are always cracks.
Today our website and API went down after a version update to a dependency our application server uses. This wasn't caught on CI because our testing application server was not a 1:1 match with the production application server (we will be changing this immediately).
All of our telemetry ingestion and alerting run totally independently, and we were the first to notify ourselves of the outage. Production monitoring more important than ever.
August Flanagan retweeted
Donald Trump starting a war in the Middle East right before midterms after inheriting an AI supercycle with a Republican House and Senate, the popular vote, and the Supreme Court
August Flanagan retweeted
The old Remington Arms factory in NY is 1.1M sq ft, closed in 2024, and sold to a developer in 2025 for $2.1M. It's in limbo, with a datacenter attempt but no real action.
Govt should buy it, do a War Assets Administration style lease to an org that bootstraps mini Shenzhen for robotics supply chain.
Does anyone else get out of the shower, check their agents, get lost in a conversation, and then realize 2+ hours later that they're standing butt naked in their living room deploying DB query optimizations? Or is it just me?
2
1
225
For me there have been three distinct phase shifts in coding in the last year.
1. Oct to Dec 2025 - holy shit these things are getting so good at doing the schlepful stuff like writing one off scripts, I can code so much faster now!
2. Jan to Aug 2026 - whoa these things are writing code as well as me. I need to stop coding myself and focus on review and test/CI infrastructure.
3. Sept 2026 to present - dang now I have a full team of these things that are better at code review and testing than me. I am now 95% product manager and 5% engineering manager.
1
2
132
Tuesday morning in my 30s: up at 5:45, gym at 6:00, get kids ready for school 7:15, leave house 8:15, drop kids at school 8:30, arrive at work 8:45.
Tuesday morning in my 40s: up at 7:15, read in bed w/ coffee until 8:15, walk dog until 9:00, surf until noon, arrive at work 1pm.
1
4
173
August Flanagan retweeted
I don’t think you have a dog in this fight but if you did you would have shot it.
I’ll take @ellalangleymsic over Kacey Musgraves any day.
And if I’m choosin’ Texas, I’ll take Ken Paxton over James Talarico.
big if true
Some rumors I've heard. I'll start with Google, then Anthropic and OpenAI.
Google: Insiders say Google has either cracked RSI (recursive self-improvement) or is very close. Their pace has picked up fast in recent months. They went from GPT-5 level to Opus 5.5 level in months, and speculation is that RSI made that possible.
Anthropic: About a week ago, Anthropic started a new pretraining run. It is not Fable 5.5 but what would become Fable 6, or something bigger, to catch up to OpenAI's Bel.
OpenAI: Bel (it may have another codename) is the model behind the math results OpenAI published a few days ago, and RSI apparently helped there too.
From what I'm hearing, OpenAI, Anthropic, and Google are all very close to cracking RSI or already have. It will get scary and exciting at the same time.
1
114
I can't believe how freaking good @bot is getting. I've been using it since a day or two after launch, but over the last week it has gotten so much more proactive. @SoniaBaschez I swear they say your tweet and turned it into an AI wife!
2
2
113
