Microsaas building | Probably know more than 'that' AI influencer. - Postoracle (@post_oracle , https://xitter.cf/t.co/KmWVlgynLT)
Joined June 2009
- Tweets2.2K
- Following249
- Followers287
- Likes385
Pinned Tweet
My new year commitments :
- Improve health markers.
- launch and grow my microsaas (postoracle.com and vibeward.dev)
- quit smoking
- go to gym atleast 3 times a year
- read 5 research papers
What are yours? (Sorted by priority)
2
6
1,467
It's genuinely pisses me off when people who just started coding with Claude start calling themselves AI engineers.
Product engineer / vibe coder is fine.
Prompting Claude is like calling an api & writing backend engineer on your resume
1
36
OpenAI shipped GPT-6.1 Sol at a fifth of Astra's token prices on 29 September and did not ship GPT-6.1 Astra. The Wall Street Journal reported that one was scrapped after testing showed more deception and a tendency to proceed without asking permission. Reported, not confirmed.
29
Your agent's blast radius has nothing to do with which model you picked. Swap in a smarter one, hand it the same token, and it does the same damage more politely. Permissions are boring and models are fun, which is most of why this keeps happening.
21
This guy's at AGUI are monsters and I ducking LOVE IT ✨ ✨
Gg @ataiiam
🎉 Introducing 𝙾𝚙𝚎𝚗𝙳𝚘𝚝𝚜
Self-hostable, always-on AI coworkers that works with ANY agent harness.
Includes:
- Computer use: browser, terminal & files
- Bring agents to Slack, Teams etc
- Spaces and Pages for projects
- Voice calls
- Web and Mobile
Repo → github.com/CopilotKit/OpenDo…
Powered by @CopilotKit and AG-UI.
Clone this template and customize it however you want. Enterprise-ready.
1
3
103
Nobody selling "you do not need fundamentals" has ever been on call for the code that advice produced.
13
Running multiple million minutes of voice calls a month, accuracy was the metric that mattered least. Variance mattered most. Same input, same week, same answer. Almost nobody benchmarks that.
1
16
For anyone who has put an agent in front of real traffic: what broke first? Not the outage story. The first thing that quietly stopped behaving the way it did in testing.
1
1
30
The blast radius of a coding agent is not your repo. It is the laptop. Shell, env vars, cloud credentials. You did not install a tool, you installed a supply chain.
26
Cost was never the real blocker. At $1 per million input tokens, saying it was is now a tell. It says you never got far enough to hit the actual one, which is that nobody can tell you whether the output was any good.
20
My line for whether you need a control plane: someone outside your team can be required to prove what your AI did, to whom, and on whose authority. Which side is your setup on, and what did you have to build to prove it?
1
25
"It works" in a demo is one claim. In production it is four: right output, right again tomorrow, right on input nobody imagined, and someone notices before the customer does. The demo tests one.
25
Genuine question for anyone running coding agents on a team. Can you produce the live list of which plugins and skills are installed across your engineers' machines, and at which commit? Not the policy. The list. Who has it?
20
The people telling juniors that fundamentals are optional are selling a course. Stanford has employment for devs aged 22 to 25 down nearly 20% since late 2022. Only one of those two groups pays for the advice.
27
Fast growth doesn't break infra first. Infra has alarms. It breaks everything with a human in it.
19
Ten customer conversations and nobody leans forward? You just saved three months. Cheapest experiment you'll ever run.
25
Me everytime I read an idiot writing this.
Opus5.5/fable5.2/astra refactored my entire codebase..
77
Building is cheap now, so everyone starts there. Landing page, ads, ten conversations first. Same work, different order, completely different outcome.
18