Teodor Bucos

28 posts

Teodor Bucos banner
Teodor Bucos

Teodor Bucos

@teobucos

👨🏻‍💻 Freelance Dev / AI Tinkerer

127.0.0.1 Katılım Kasım 2025
287 Takip Edilen9 Takipçiler
Teodor Bucos retweetledi
sidbing 🪽
sidbing 🪽@sidbing·
ingest and read everything you need to know, docs, code etc. step away from ur 'puter, sit with a pen and paper and write stuff down until feels complete. when facing ambiguity, ask the agent/read some more and step away again. until what you have on the paper feels solid. then wispr flow into an agent and ask it to keep asking u questions until it's also on the same page. then hit /goal with the spec.md and step away
Matt Stockton@mstockton

Stepping away from the dopamine chat box and recording a long, rambling, yet intentional voice memo covering everything you want to accomplish in whatever project you’re working on. Then transcribing that and using it as input to the model, where you then ask the model to ask you questions exhaustively until it knows exactly what to do in order to accomplish your goal

English
0
3
20
1.5K
Teodor Bucos
Teodor Bucos@teobucos·
Social media is a big echo chamber. You get excited about a new idea, research it, find almost no competition, start building... then suddenly everyone seems to be solving the same problem. Maybe the zeitgeist just ships faster than I do.
English
0
0
0
9
Teodor Bucos
Teodor Bucos@teobucos·
Hey @thsottiaux, any chance of bringing back the context usage indicator in Codex? It was very helpful for understanding the weight of each additional message in a chat before manually deciding to compact.
English
0
0
0
49
Teodor Bucos
Teodor Bucos@teobucos·
Also, credit where it’s due: from my own experience, Amp is by far the #1 coding harness out there. Kudos to the team 👏🏼
English
2
0
2
150
Teodor Bucos
Teodor Bucos@teobucos·
$819 later, @ampcode discovered I treat my AI usage like a challenge 😆 I'm trying to tokenmaxx all my subs and usage every day. I knew the $15 a day wasn’t going to be around for long after they stopped ads, so I tried to make the most of it. Completely understand and thank you for the very generous run 🙏
Teodor Bucos tweet media
English
1
0
6
396
Teodor Bucos
Teodor Bucos@teobucos·
Capabilities will be the unit of agentic commerce, the same way attention is the unit of human commerce.
English
0
0
0
16
Teodor Bucos
Teodor Bucos@teobucos·
Machines of Loving Grace. You can see it when reading Opus reasoning traces. And it just feels like magic.
English
0
0
0
15
Teodor Bucos
Teodor Bucos@teobucos·
GPT 5.5 inside Codex seems slower and dumber today. Anyone else noticed any difference? I'll give Grok a try in @opencode
English
1
0
1
35
Teodor Bucos
Teodor Bucos@teobucos·
I’ve been a Windows user my whole life. As I started getting deeper into development, the experience started to feel more and more frustrating. So I decided to give Mac a try. The first few days were rough: new shortcuts, new settings, lower productivity. Then I set up Raycast, and things slowly started to feel like magic. One week in, I get it. The productivity is just on another level. Not going back.
English
0
0
1
25
Teodor Bucos retweetledi
Yohei
Yohei@yoheinakajima·
i'm excited to open source Active Graph: an event-sourced reactive graph runtime for long-running, agents 🔄🧠 events/logs projects a graph. reactive behaviors react and affect the graph. fork-and-diff agent runs. no A2A, no workflows, no DAG site: activegraph.ai docs: docs.activegraph.ai github: github.com/yoheinakajima/… quick start: pip install activegraph this is an early experiment in a new paradigm for agent architecture 🧪
English
55
52
524
90.1K
Cloudflare
Cloudflare@Cloudflare·
Cloudflare's security team spent the last few weeks testing Anthropic's Mythos against fifty of our own repositories. What we learned about offensive AI, why faster patching is the wrong reaction, and what the architecture around vulnerabilities has to look like next. cfl.re/49BRUqW
English
85
713
4K
1.6M
Teodor Bucos
Teodor Bucos@teobucos·
The takeaway: Security teams can’t rely on patch speed alone anymore. When attacks are automated, rushing fixes under impossible deadlines can create even bigger problems. The better strategy is to build systems where vulnerabilities are harder to reach, harder to chain together, and harder to turn into real attacks.
Teodor Bucos tweet media
English
0
0
1
75
Teodor Bucos
Teodor Bucos@teobucos·
The most critical part of the harness is the "Trace" stage. The AI checks whether a real attacker could actually reach and exploit the bug from outside the system. This is the proof that separates a harmless flaw from a real vulnerability.
Teodor Bucos tweet media
English
1
0
0
82
Teodor Bucos
Teodor Bucos@teobucos·
@AnthropicAI Mythos model was tested by @Cloudflare on more than 50 real repositories and it didn’t just find bugs. What they discovered about cyber frontier models is a wake-up call for every software company. Thread 🧵
Cloudflare@Cloudflare

Cloudflare's security team spent the last few weeks testing Anthropic's Mythos against fifty of our own repositories. What we learned about offensive AI, why faster patching is the wrong reaction, and what the architecture around vulnerabilities has to look like next. cfl.re/49BRUqW

English
1
0
1
1.4K