Caleb Shepherd

801 posts

Caleb Shepherd banner
Caleb Shepherd

Caleb Shepherd

@caleb_shepherd

AI researcher. And by researcher I mean stalking AI accounts on X all day. Follow for AI research papers that most people miss.

Los Angeles, CA Katılım Eylül 2017
201 Takip Edilen31 Takipçiler
Caleb Shepherd
Caleb Shepherd@caleb_shepherd·
I just built a tool that will save companies over $100,000 a year on labor. More on this soon.
English
1
0
0
13
Caleb Shepherd
Caleb Shepherd@caleb_shepherd·
@thsottiaux I trust when it breaks my expectations of what it can do. The first time I felt that was with gpt-5.3-codex and felt it again with 5.5. The other guys have always disappointed me, Codex has yet to.
English
0
0
6
589
Tibo
Tibo@thsottiaux·
Do you still trust benchmarks or do you just listen to your friends? What makes you try a new model?
English
857
31
1.8K
148.5K
Arm
Arm@Arm·
A new era of PC. 25.0528, 121.5990
English
345
658
10.4K
1.8M
Caleb Shepherd
Caleb Shepherd@caleb_shepherd·
GPT-5.6 rumors are heating up. Massive context. Agentic reasoning. Codex-native workflows. Persistent memory. Multimodal orchestration. Autonomous debugging. Long-horizon execution. But the feature everyone really wants: A persistent world-model workspace. An AI that remembers your repo, docs, roadmap, bugs, design system, and product decisions — then builds like an always-on technical cofounder. No re-explaining. No context resets. No babysitting. If GPT-5.6 gets close to this, it’s not a chatbot upgrade. It’s ambient intelligence going live.
English
1
0
0
161
Caleb Shepherd
Caleb Shepherd@caleb_shepherd·
o3 should have been GPT-5
English
0
0
0
17
Vaibhav (VB) Srivastav
codex updates rolling out now: - computer use on windows - remote control windows hosts from chatgpt mobile or mac - new profile with token stats, streaks, longest task + more - performance improvements and bug fixes p.s. this tweet was posted via codex mobile, controlling browser on windows
Vaibhav (VB) Srivastav tweet media
English
46
14
346
25.2K
Caleb Shepherd
Caleb Shepherd@caleb_shepherd·
@kimmonismus Yes, I've been saying this. The GPT-5 release seemed so boring but only because it was technically rolled out incrementally. I remember right after GPT-4 came out Sam said GPT-5 would be rolled out incrementally instead of one big release, and he was right.
English
0
0
1
159
OpenAI
OpenAI@OpenAI·
Windows users, this one’s for you. Computer use now works on Windows, so Codex can take action on your Windows computer. And with Windows support for Codex in the ChatGPT mobile app, you can start, review, and steer tasks on the go while work continues on your Windows machine. An early experience, but we’re working on more ways to keep your work moving, wherever you are.
English
755
840
7.9K
1.1M
Caleb Shepherd retweetledi
Tibo
Tibo@thsottiaux·
Codex Thursday has exceptionally moved to another day. Friday it is.
English
483
130
4.8K
451.6K
Caleb Shepherd
Caleb Shepherd@caleb_shepherd·
@Angaisb_ Vision is buttcheeks right now. I really hope they focus on that
English
0
0
0
17
Angel 🌼
Angel 🌼@Angaisb_·
I can't wait for the models to be able to play video games in real time, we're slowly getting there, computer use is getting faster with every model release The only thing that hasn't improved much is vision
English
6
1
57
4.2K
Caleb Shepherd
Caleb Shepherd@caleb_shepherd·
@LexnLin I'll wait until the community reviews it, it's meaningless coming from someone who works at Anthropic
English
0
0
0
66
Caleb Shepherd
Caleb Shepherd@caleb_shepherd·
@BayernHahn @LexnLin It was pretty sizable, definitely not a small leap. The more saturated a benchmark gets, the more meaningful the little percentage jumps get. 30% to 40% is not as meaningful as 70% to 76%.
English
1
0
1
15
Hahn
Hahn@BayernHahn·
@LexnLin But the Leap in the Benchmarks wasnt that huge or am i Missing Something
English
1
0
0
67
Caleb Shepherd retweetledi
Claude
Claude@claudeai·
Introducing Claude Opus 4.8: it builds on Opus 4.7 with sharper judgment, more honesty about its own progress, and the ability to work independently for longer than its predecessors. Available today at the same price.
Claude tweet media
English
3.5K
8.6K
66.8K
14.3M
Caleb Shepherd
Caleb Shepherd@caleb_shepherd·
@daniel_mac8 Exactly. There have been no hints that it will release today. Only hints at Codex.
English
1
0
2
1.8K
Dan McAteer
Dan McAteer@daniel_mac8·
Unpopular opinion: GPT-5.6 will not be released today. It will be Codex upgrades.
English
39
4
643
56.4K
Caleb Shepherd
Caleb Shepherd@caleb_shepherd·
Hold your horses people, 5.6 is not dropping today
English
0
0
0
17