ben

216 posts

ben banner
ben

ben

@CalmCoding

FAANG SWE | Shipping AI Agents & Code That Works in the era of slop | Finance enjoyer | Memes when they're the most efficient explanation

New York, NY शामिल हुए Mart 2026
66 फ़ॉलोइंग26 फ़ॉलोवर्स
पिन किया गया ट्वीट
ben
ben@CalmCoding·
Is it just me or are custom SKILL.md files by far the most useful AI productivity trick right now? I’ve never had success with whatever current thing people were raving about, like Ralph loops, openclaw (although I do see the potential here and have tried it but couldn’t find a use for me yet), multi-agent setups etc. Like my steps are: 1.Identify something painfully repetitive and annoying to do that you do all the time 2.Put the exact steps in a skill file and load it in the agent 3.Profit But idk maybe this is a skill issue for me.
English
1
0
6
484
ben
ben@CalmCoding·
@tunguz I’m not laughing. But they’ve been saying this for a couple years now and I still have my agent catastrophically forget information within context.
English
0
0
4
534
Bojan Tunguz
Bojan Tunguz@tunguz·
You are laughing. Superhuman AI has just been confirmed and you are laughing.
English
38
13
295
16.5K
ben
ben@CalmCoding·
@andrewchen Sounds like a take home project instead you now have to go to the office instead of the comfort of your own home.
English
0
0
4
298
andrew chen
andrew chen@andrewchen·
noticing a trend of startups replacing standard resumes/interviews with week-long (or at least 3-day weekend) in-office trials. Makes sense in a world of AI-generated resumes and interview responses Turns out the best signal for whether someone can do a job is watching them actually do the job. took us 100 years of HR to rediscover apprenticeships!!! 😂
English
139
55
943
111.7K
ben
ben@CalmCoding·
@creatine_cycle If you are not using mythos already you are ngmi
English
0
0
1
65
atlas
atlas@creatine_cycle·
any startups vibecoded before claude mythos are over. any startups vibecoded after claude mythos will stand the test of time
English
10
3
107
3.2K
ben
ben@CalmCoding·
@Doomerzoomer For now perhaps. But there was a time OpenAI, Microsoft were in the drivers seat. Things can change quickly
English
0
0
0
1.9K
ben
ben@CalmCoding·
@icanvardar The entire release is marketing copy
English
0
0
1
90
ben
ben@CalmCoding·
@RayFernando1337 Has anyone considered the reasons behind these restrictions has a lot to do with marketing. Kinda brilliant ngl
English
5
0
13
4.7K
Ray Fernando
Ray Fernando@RayFernando1337·
Project Glasswing FAQ: Q: Why only 12 companies? A: They're the ones who can afford us. Q: What about open-source maintainers? A: We found bugs in their code. You're welcome. Q: Will you release the tool publicly? A: We said "cybersecurity is the security of our society." We didn't say which society.
Anthropic@AnthropicAI

Introducing Project Glasswing: an urgent initiative to help secure the world’s most critical software. It’s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans. anthropic.com/glasswing

English
35
50
871
192.8K
ben
ben@CalmCoding·
@kjnormac @_ontologic Yes and they make it out like it’s a skill issue on our part.
English
0
0
0
19
Beast With Shades On
Beast With Shades On@kjnormac·
@CalmCoding @_ontologic Fucking thank you. Normal people in the AI dev space know this for a fact but these AI proselytizers are acting like it's not a problem.
English
1
0
1
49
∿spencer.​​​​​​​​​​​​​​​​​​​​​​​​​​​​​
∿spencer.​​​​​​​​​​​​​​​​​​​​​​​​​​​​​@_ontologic·
> LLMs have hit a plateau Even if this were true, they’re capable of an insane amount of work already. For most workflows hallucinations don’t matter nearly as much as everyone seems to think
Ewan Morrison@MrEwanMorrison

People are waking up to the fact that LLM AIs have hallucinations baked in. They can't be fixed. LLMs have hit a plateau, should not be used for serious work. The vast data centre build is a pointless waste of money & resources just to double-down on a mistaken assumption.

English
19
9
122
21.8K
ben
ben@CalmCoding·
@chooserich It's time for dueling AIs
English
0
0
1
109
Nick O’Neill
Nick O’Neill@chooserich·
This is D-Day for software companies. It's also a hostage situation. If large software companies don't pay Anthropic for their new cybersecurity model THERE IS AN 85% CHANCE THEY WILL BE HACKED. This isn't innovation, it's a shakedown.
Anthropic@AnthropicAI

Introducing Project Glasswing: an urgent initiative to help secure the world’s most critical software. It’s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans. anthropic.com/glasswing

English
81
31
490
58.5K
Heisenberg
Heisenberg@Mr_Derivatives·
Every 10 minute is a new headline… Just wake me up when all of this is over…
English
92
33
1.2K
50.6K
ben
ben@CalmCoding·
@amritwt That we can’t even use
English
0
0
0
34
ben
ben@CalmCoding·
@TheGeorgePu Honestly kinda messed up. Did we just accept that they can freely scrape anything off the internet?
English
0
0
2
76
George Pu
George Pu@TheGeorgePu·
Anthropic just released its most powerful AI model yet. Except we can't use it. Only Amazon, Apple, Google get access. Trained on millions of our conversations. We built it for them.
George Pu tweet media
English
16
3
36
1.8K
ben
ben@CalmCoding·
@almmaasoglu What benefit does an engineer have personally for doing this? They can't use the models for personal use
English
0
0
0
41
Alim
Alim@almmaasoglu·
Calling it now engs will join orgz just to be able to use big models that will be gatekept from public APIs. Screencap this
English
1
0
11
456
ben
ben@CalmCoding·
@PaulSkallas It went to TV for a bit and now it's nowhere
English
0
0
0
2.3K
LindyMan
LindyMan@PaulSkallas·
They stopped making good movies. Can't believe it. imagine telling someone this 20 years ago. "they don't make good movies in the future anymore" what?
English
171
150
3.8K
99.4K
ben
ben@CalmCoding·
@AIExplainedYT Security through obscurity generally not a good practice. Open source will catch up eventually
English
0
0
0
1.3K
ben
ben@CalmCoding·
@TrentonBricken Pretty sure they've said things like this before, Anthropic likes to be dramatic.
English
0
0
1
698
Trenton Bricken
Trenton Bricken@TrentonBricken·
Mythos sandbox escape and many more wild instances are in the Model Card
Trenton Bricken tweet media
English
11
23
270
32.9K
ben
ben@CalmCoding·
@kipperrii At this point it's going to be battling AIs in cybersecurity. Straight out of SciFi
English
0
0
0
106
kipply
kipply@kipperrii·
we'll try our best but there's a pretty good chance that we (everyone) fail to do controls well enough as to not make the state of cybersecurity unrecognisable
Anthropic@AnthropicAI

Introducing Project Glasswing: an urgent initiative to help secure the world’s most critical software. It’s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans. anthropic.com/glasswing

English
3
1
55
3.4K
ben
ben@CalmCoding·
@kimmonismus It's easy to game the benchmarks, but no doubt this is a leap anyways. We'll have to see how it behaves in practice.
English
0
0
3
1.3K
ben
ben@CalmCoding·
@Zai_org How well could these open source models perform if they didn’t distill from closed source
English
0
0
0
1.8K
Z.ai
Z.ai@Zai_org·
Introducing GLM-5.1: The Next Level of Open Source - Top-Tier Performance: #1 in open source and #3 globally across SWE-Bench Pro, Terminal-Bench, and NL2Repo. - Built for Long-Horizon Tasks: Runs autonomously for 8 hours, refining strategies through thousands of iterations. Blog: z.ai/blog/glm-5.1 Weights: huggingface.co/zai-org/GLM-5.1 API: docs.z.ai/guides/llm/glm… Coding Plan: z.ai/subscribe Coming to chat.z.ai in the next few days.
Z.ai tweet media
English
436
1.2K
9.4K
2.9M
Garry Tan
Garry Tan@garrytan·
I’m on a United flight with their legendarily bad WiFi that makes internet unusable on my MacBook Pro especially for heavy multi-worktree @conductor_build sessions But now I have a new love: building in multiple topics in Telegram group with my OpenClaw bot It feels amazing
Garry Tan tweet media
English
52
17
268
57.6K

खोजें

@tunguz @andrewchen @creatine_cycle @Doomerzoomer @icanvardar @RayFernando1337 @kjnormac @_ontologic