Sabar

I build things that think.

Engineer by profession. Pattern hunter by instinct.

[Sabar @SabarnathX]

@nicbstme Wait till you hear about copilot studio and ai foundry business adoption.

[Sabar @SabarnathX]

petabytes of agent logs. 1 PB = 1000 TB

an engineer’s perfect nightmare

would love to know how they’re doing that at this scale

[Sam Altman @sama]

There is an extensive and ongoing review related to our agents’ use of internet access during training and evaluation. We’ve been publishing summaries at the link below and will continue to.

We have not been as fast as we would have liked but we are trying to balance our desire for transparency with gaining a clear understanding from petabytes of agent activity logs, and working with impacted organizations.

We are prioritizing as best as we can based on severity, and adding resources. Hugging Face is still the most severe event we’ve seen. We will be as transparent as we can be subject to things like vulnerabilities in other companies that our agents have found, which will be their call to disclose or not.

[Sabar @SabarnathX]

Giving an agent autonomy is easy. Defining “done” is the actual work.

Otherwise, autonomy is just unsupervised confidence.

[Sabar @SabarnathX]

Stop asking whether the new model is smarter.

Ask which failure it removed - and which new one it introduced.

[Sabar @SabarnathX]

AI software factories can generate code faster than most teams can review it.

The bottleneck did not disappear. It moved from writing code to deciding what deserves to ship.

[Sabar @SabarnathX]

Is prompt engineering dead?

[Sabar @SabarnathX]

@ns123abc Congratulations, Nik! This is well deserved.

[Sabar @SabarnathX]

Fluency wins voice - agent demos. Recovery wins trust.

Interrupt it, break a tool, or change the task mid-sentence - then judge it.

[Sabar @SabarnathX]

We are rapidly moving production from purely deterministic systems toward probabilistic ones.

[Sabar @SabarnathX]

Longer context windows are becoming an excuse for worse context management.

Good context is curated, not accumulated.

[MTS @MTSlive]

Hacktron founder @S1r1u5_ reveals how his team hacked into OpenAI starting with a single HEIC image and ended with Codex opening a PR under an employee’s token:

"The main one was an SSO misconfiguration in OpenAI's identity system. The other bug is a remote code execution on community.openai."

"By just uploading an HEIC image, an image format that you might have noticed in iPhone, we were able to get remote code execution there."

"Every employee that logins to Discourse, we can use their token and use it on Codex."

"We asked Codex to update the README with, Hacktron AI team is here. Don't change anything else. And then we clicked on open PR, which led to opening a PR in OpenAI."

"When you create Codex environments, you can give your own Bash scripts. You can write a Bash script that can exfil the source code of those internal repositories. You don't even need to ask Codex to interact with the repo."

"We can technically do this with every single repository that was on github .com/openai."
@HacktronAI @rootxharsh

[Sabar @SabarnathX]

@NyrVani Really enjoyed this

sharp, funny, and the bedtime spiral feels completely real.

[Sabar @SabarnathX]

Finally AGENTS . md is the shared standard

[Sabar @SabarnathX]

Not a solution, but a workaround until Claude fixes this.

[Sabar @SabarnathX]

The most dangerous AI-generated code is clean, plausible, and built on a decision nobody approved.

Review the assumptions first. Syntax can wait.

[Sabar @SabarnathX]

Interesting OpenRouter pricing observation:

GLM-5.3 was listed at $0.8775/M input and $2.97/M output, but my requests were routed to Wafer at $1.40/M and $4.40/M.

The billing math was correct for Wafer. At the displayed rate, the same usage would have cost ~$13.66 instead of $20.36 - roughly 49% less.

Check the actual provider in your logs and enable price-based routing (:floor or provider.sort: "price") when cost is the priority.

Takeaway: the headline model price is not guaranteed.

@OpenRouter

[Sabar @SabarnathX]

@Cch_Chichieh 💯

[Sabar @SabarnathX]

Recently I'm having trust issues with Codex.

It always sneaks into my other projects and workspaces.

sandboxMaxxing

[Sabar @SabarnathX]

Beautiful Pichwai Art

@TiffanyATrump pure grace 💗

[Tiffany Ariana Trump @TiffanyATrump]

India 💗

[Sabar @SabarnathX]

[Task]
[Context]
[Expected Result]

---

Let the AI do the rest 🪄

[Sabar @SabarnathX]

Long context is not memory.

It is a larger room where stale assumptions and missing decisions are harder to notice.

A useful handoff preserves decisions, evidence, and unresolved risks - not conversational filler.