Michael Ramos

1.2K posts

Michael Ramos

Michael Ramos

@backnotprop

Cofounder, AI @EQTYLab / prev dc - complex systems / veteran / For fun: @plannotator

CA Katılım Mart 2024
304 Takip Edilen904 Takipçiler
Sabitlenmiş Tweet
Michael Ramos
Michael Ramos@backnotprop·
Anthropic code review this, clanker review that ... why don't you shut up and review+annotate your own code.... (yes im a loser who still manually reviews code) Originally inspired by a bunch of feature requests and then seeing @dillon_mulroy tweet a similar cool ux. @plannotator for reviewing plans (primary focus) and code, fully oss. OpenCode, @badlogicgames 's pi.dev, and Claude Code and other clankers
English
14
17
279
63.2K
Michael Ramos
Michael Ramos@backnotprop·
@dexhorthy You can outsource the thinking, it just may be wrong and misaligned
English
0
0
0
3
Jamon
Jamon@jamonholmgren·
Unpopular opinion: 1- or 2-letter variable names in focused, obvious contexts are totally fine.
Jamon tweet media
English
105
7
256
431.6K
Michael Ramos
Michael Ramos@backnotprop·
@petergyang Now the kids can go out and socialize vs stay up all night at shack15
English
0
0
1
332
Peter Yang
Peter Yang@petergyang·
How do people even do AI hackathons these days you're just sitting around waiting for the agents half of the time?
English
141
8
680
76.7K
Michael Ramos
Michael Ramos@backnotprop·
The more i use `/goal` the more I want to ensure the model gets simple facts right up front. So I use a version of @mattpocockuk grill-me focused on @Everlier's methodology around facts. Facts allow me to describe very clear requirements for a feature or system. & Not always that technical. The more I do with AI, the less patient I get, so I can't read a ton of markdown, but I can afford to read/verify simple lists of facts. Once facts are established i let the agent refine a plan for itself (every decision aligned to those facts) so it can figure out order of operations. have been having a lot of fun with this process. composition by @editframe
English
9
9
156
11.1K
Michael Ramos
Michael Ramos@backnotprop·
I'm lazy and rely on the model creating the facts. But I try to make it clear just in the first statement. /setup-goal "Everything I want to make clear." & then rely on the QA grill me phase to ensure alignment. Model crease loose effects. I'm creating a UI right now to make the process even smoother.
English
0
0
3
129
BruzWJ
BruzWJ@BruzWJ·
@backnotprop @mattpocockuk @Everlier when i tried this, the part i actually trusted was the 'verify simple lists of facts' step, not the long markdown. do you have the model draft the list and grill it, or write the facts yourself first?
English
2
0
1
178
Everlier
Everlier@Everlier·
Thanks for the Facts shout-out! This workflow was one of the things I wanted to acheive - having an arbitrarily defined fact sheet and then work on an implementation until all are done. I like how you're combining the fact sheet with /goal and /grill-me, gives me a few ideas for upcoming versions
English
1
0
2
299
Michael Ramos
Michael Ramos@backnotprop·
IMO this will require Anthropic to create a moat around better models. We as an industry have remained fairly adverse to lockin.
English
0
0
1
60
Michael Ramos
Michael Ramos@backnotprop·
It’s interesting to watch a company go very deep into the Oracle strategy in 2026. vs Google/Kubernetes - which we are sort of seeing play out in a short time frame with Codex’s strategy remaining open source and integration friendly.
English
1
0
8
570
dax
dax@thdxr·
we are impacted by bun stability more than almost anyone, anything they're doing to try and improve that is good simultaneously we're going to wait a long time before considering moving a million daily users to the rust version so we're stuck for now
English
104
27
2.2K
154.8K
Michael Ramos
Michael Ramos@backnotprop·
Ending free lunch is fine. Understandable. But why not enable an open ecosystem to build around your intelligence capabilities. This is the opportunity codex has - similar to Kubernetes effect for agentic infrastructure. …While it remains open source and integratabtle in loose ways.
English
0
0
3
204
Old Man Jake
Old Man Jake@oldwhippasnappa·
@backnotprop Yes! At this rate, you'll be writing an editor :P fyi, tried the latest version of plannatator -- am a fan! Use it most days at work. Thank you!
English
1
0
0
12
Michael Ramos
Michael Ramos@backnotprop·
planned out and ran a low stakes overnight goal to create psuedo-lsp capabilities, should I ship it?
English
6
1
15
3K
Michael Ramos
Michael Ramos@backnotprop·
@theo Ergonomics and performance. Shell capabilities/shortcuts. Ghostty + macos gives me everything i need in terms of dynamic navigation and windowing. Maybe I've had one pain of trying to find a historic session (ie tab sidebar).
English
1
0
20
967
Josh
Josh@joshmanders·
@backnotprop @ridafkih Yeah man heaven forbid a company makes money from their own service that you're using to make money for yourself. So absolutely selfish of them.
English
2
0
5
103
Josh
Josh@joshmanders·
I find this reasonable. Pro/Max accounts should be exclusively for human usage. Use the API if you want automation, and pay appropriately for it. But also, if you don't want automation, remove the -p flag from Claude Code instead of this weird dance.
ClaudeDevs@ClaudeDevs

Starting June 15, paid Claude plans can claim a dedicated monthly credit for programmatic usage. The credit covers usage of: - Claude Agent SDK - claude -p - Claude Code GitHub Actions - Third-party apps built on the Agent SDK

English
35
3
120
16.9K
Josh
Josh@joshmanders·
@backnotprop @ridafkih They can switch to API billing and continue on exactly as they were. The only change her is you no longer get unlimited token spend for $200/mo for any automated stuff.
English
1
0
1
114
Michael Ramos
Michael Ramos@backnotprop·
@joshmanders @ridafkih eh. This neglects users who built development workflows around programmatic usage in their own environments. It's their product in the end. That's the only justifiable thing here.
English
1
0
1
97
Josh
Josh@joshmanders·
@ridafkih I see it as plugging a hole where people were abusing their subs to circumvent api pricing.
English
3
0
4
443
jaden
jaden@JadenGeller·
@backnotprop @theo it’s new separate pool of usage credit, but it’s a much smaller pool than was available before for those tools on the sub
English
1
0
1
67
Theo - t3.gg
Theo - t3.gg@theo·
If you use any of the following with your Claude sub, your usage must got cut by 25x: - T3 Code - Conductor - zed - jean - “Claude -p” in your ci - scripts to call Claude code from other tools They’re disguising this as “free credits”. Don’t fall for it.
ClaudeDevs@ClaudeDevs

Starting June 15, paid Claude plans can claim a dedicated monthly credit for programmatic usage. The credit covers usage of: - Claude Agent SDK - claude -p - Claude Code GitHub Actions - Third-party apps built on the Agent SDK

English
201
166
3.3K
365.1K

Keşfet

@dexhorthy @jamonholmgren @petergyang @mattpocockuk @Everlier @editframe @thdxr @elonmusk