Ex-Riot Games. One role I hired for: 1,200 resumes, 42 screens, 2 finalists. I show people in games how to be one of the two. Free weekly newsletter ↓
Singapore
Joined March 2023
- Tweets4.2K
- Following1.7K
- Followers682
- Likes511
Pinned Tweet
Most career advice is garbage.
"Follow your passion."
"Just work hard."
"Your time will come."
That's how smart people get stuck for 10 years.
Let me show you how to play a different game. 👇
Asking an AI to "fix everything" in a project can come back as one giant change that's hard to review.
@alexalbert__ of @AnthropicAI asks for a split instead, on every project, each time a new model ships: "Fix what you're sure about and list the rest for me to review."
Two outputs: one you can accept, one you can judge.
teapawt.org/alexalbert__/status/21…
Every time we have a new model, I run the same simple prompt on all my projects:
"Read this entire project and find everything that’s broken, inconsistent, inefficient, or no longer needed. Fix what you’re sure about and list the rest for me to review."
highly recommend trying
17
Hiring signal from @OpenAI: Researcher, Alignment CoT Monitorability, San Francisco (listing checked 11 Nov).
The requirement, in their words: "move from an ambiguous model-behavior question to a concrete experimental setup" and "decide what the evidence supports."
A small artifact that tests it: one hypothesis about an open model's reasoning, a short eval, the result, and one thing the evidence does not support.
Apply on the official listing only.
jobs.ashbyhq.com/openai/8249…
23
If AI now does most of the grunt work juniors used to learn on, where will the next generation of senior hires get their reps? Name one concrete path a 22 year old should take this year. @grok
1
35
If the recruiter and the candidate both send AI agents to negotiate the salary, who comes out ahead, and which part of that negotiation should the human still handle themselves? @grok
1
21
Claude Fable 5.2 is out, and @AnthropicAI's launch page still tells most users to start on Opus 5.5.
Fable is for the most ambitious, open-ended technical work.
Test it on one hard task, not your whole workflow.
anthropic.com/claude-fable-a…
1
1
85
Ok, so why is the new Fable and Mythos number "5.2" and not "5.5????"
@AnthropicWay, are you still holding back on us??? :)
1
26
in case people ever were curious as to why T3 is hands down the best...
looking forward to the GTA6 integrations. hahaha
After @BellularGaming's take on a native @t3dotcodes addon for World of Warcraft instead of my overlay approach, I had to take a swing.
This is better in every way. It feels totally native, and with full controller and dictation support I can drive threads from the couch.
51
Which game studio job will AI change the most by 2028: QA tester, localization, concept artist, or community manager? Pick one and tell that person what to learn now. @grok
2
52
US tech unemployment fell to 2.8% in October (@CompTIA, Nov 6).
The AI detail is more useful: active postings asking for AI skills rose 6.2% to 366,322, and dedicated AI job titles grew 9.8%. CompTIA counts them separately because the skills number includes jobs that may not carry an AI title.
So search postings for the skill, not just the title.
prnewswire.com/news-releases…
20
A good father teaches his children how to dream, let's the experience what it is to fly and catches them when they fall.
I'm learning every day.
I believe kids need room to make mistakes and learn to handle a little discomfort. A father can still be there to guide them.
That thought inspired First Flight, my little AI film about a boy who wants to fly.
Always a thanks to @Artedeingenio for the creative direction that helped bring my idea to life.
28
"Our agent ships features" is a common pitch. Here's the bar I'd want to see:
One real feature request. The agent's pull request. Every acceptance check written before the work started, each marked pass or fail. And how long a human took to review it.
Show me that and I'll believe it.
28
Using Claude Code from @AnthropicAI? Check your repo for .claude/settings.local.json.
@ApiiroSecurity found 1,122 public GitHub copies of that file holding at least one secret. Picking "Yes, and don't ask again" on an exact command can save it word for word, password included.
Deleting the file isn't enough, since git history and forks keep it. Rotate the keys first, then clean up.
thehackernews.com/2026/11/cl…
33
A role I think is coming fast: the agent reliability lead.
@honeycombio just put SRE agents in early access where each finding links to the query, trace, or deploy behind it, says whether it's confirmed or a hypothesis, and gets more autonomy only as it earns it.
The person running that can:
- set the goals agents work toward
- read an evidence trail and catch a bad call
- decide what an agent may do alone
- turn repeat misses into runbooks
That's a skill set before it's a job title.
honeycomb.io/blog/honeycomb-…
1
2
95
I run 12 Grok Bots daily. Three asks:
1. Let me start a fresh chat with an existing bot and keep its memory, files and routines
2. Tell us if usage per message grows as a chat gets longer, before compaction
3. Show usage per message or per run, not just by day
Long chats are my biggest cost and I can't see why. Any chance @poteto?
1
18
Building something with AI this weekend? Reply with the link and one sentence on what you were trying to prove.
I'll leave a short note on what's clear and what's missing, the same way I'd read a work sample.
One project, one goal, and one thing you'd change next.
15
If you work in SEO or content, ChatGPT appears to be testing a dashboard where verified businesses can see how often they show up in answers, and how many people click through. @OpenAI has not announced it (Search Engine Land, Nov 6).
It reportedly will not show the prompts that caused each mention.
A useful habit now: keep a private log of queries where your work should appear, so you have a baseline before any official charts exist.
searchengineland.com/openai-…
1
26
Ground zero for the slopocalypse
Roblox latest earnings call on how gamers using AI:
“Over the past month, more than 80% of the top 100 games on Roblox (ranked by hours engaged) used AI tools to accelerate creation, and the number of daily asset uploads to the Roblox cloud tripled month-over-month.
Since Q2, time spent using AI tools (including Build, Studio Assistant, and external AI via MCP) tripled among our top creators, and those using our MCP server daily grew nearly 5x.”
22
If you're hunting a junior AI role this week, AI Native Careers' entry-level filter lists 1,360 openings across 83 companies this morning (out of 14,849 roles at 133 firms).
Some titles on page one still read senior, so open the posting before you assume it's junior.
Apply only if the actual listing fits, and attach one artifact a stranger can judge:
ainativecareers.com/ai-jobs/…
20