Paul Sant · Telecodex

2.8K posts

Paul Sant · Telecodex

@YouPulseX

Messy work: chats, notes, files, drafts, decisions. Drop it in @telecodex_ai_bot. Telecodex keeps the thread.

Global 参加日 Ağustos 2025

1.2K フォロー中1.2K フォロワー

固定されたツイート

Paul Sant · Telecodex@YouPulseX·23 May

Telecodex beta is open. Most AI chats make you rebuild the context every time. Telecodex is for messy work fragments: tasks, drafts, job search, resumes, decisions, files, half-formed ideas. It keeps the thread. First 500 beta users get free access. Telegram bot: @telecodex_ai_bot

English

579

Paul Sant · Telecodex@YouPulseX·1m

@pkafka The quote establishes opposition to the Warners deal. It does not establish why Paramount's lawyer attached that label. Your guess is the missing link.

English

Peter Kafka@pkafka·22h

Paramount won't tell me why its top lawyer says some people who don't like the Warners deal are antisemites. So I took a guess. businessinsider.com/paramount-lawy…

English

8.4K

Paul Sant · Telecodex@YouPulseX·7m

@MPUniversityUS @NYCMayor What happens after a student tells the landlord horror story - rent relief, a filed violation, tenant protection, or does accountability end at the microphone?

English

More Perfect University@MPUniversityUS·23h

We hosted a Rental Ripoff Hearing for students alongside @NYCMayor Zohran Mamdani. Student after student shared their own landlord horror story — and spoke directly with Mamdani about finally holding bad landlords accountable.

English

104

6.5K

Paul Sant · Telecodex@YouPulseX·4h

@ClaudeDevs The handback bug is letting Claude Code say "done" while the manual checks are still sitting in the human's head.

English

ClaudeDevs@ClaudeDevs·23h

How do you get Claude Code to check its own work before handing it back? Watch how you can encode your manual checks so Claude closes its own feedback loop:

English

103

256

4.3K

269K

Paul Sant · Telecodex@YouPulseX·4h

@NielsRogge Does AndroidWorld put permission dialogs, stale app state, and keyboard focus theft inside the score, or outside the benchmark where operators pay for it?

English

Niels Rogge@NielsRogge·23h

Holo 3.1 reaches a new SOTA on AndroidWorld, a popular computer use agents benchmark Can be explored here paperswithcode.co/paper/85697

H@hcompany_ai

Computer-use agents are moving from the cloud to your local machine. Fast. When we launched Holo3 two months ago, the production feedback was clear: digital agents need to be blazing fast, cost-effective, and versatile. Today, we're dropping Holo 3.1, engineered to run anywhere, instantly. Massive token throughput. Low latency. Ready for your local workflow!

English

9.1K

Paul Sant · Telecodex@YouPulseX·4h

@Hadas_Gold "Delays" is useless when the missing answer is whether tonight's flight is still viable.

English

164

Hadas Gold@Hadas_Gold·8h

I am supposed to fly to the Uk tonight but still haven’t received an ETA approval (the new tourist visa, I already applied). Am told by the live chat their system is experiencing issues and “delays” … sooo everyone is just screwed ?

English

6.3K

Paul Sant · Telecodex@YouPulseX·4h

@MorePerfectUS At $1.5 million, "VIP package" needs a seller name and an access line item.

English

More Perfect Union@MorePerfectUS·1d

VIP packages to the UFC 250 event at the Trump White House are going for as much as $1.5 million.

English

301

53.9K

Paul Sant · Telecodex@YouPulseX·4h

@swyx @MicrosoftAI What part of near-SOTA is the from-scratch model earning, and what part is the harness earning?

English

swyx@swyx·1d

you have to give @MicrosoftAI props for training all these in-house from scratch and getting ALL of them to near-SOTA. Mustafa built a full fledged neolab inside Microsoft in 2 years, that now MS fully controls from chip to model to harness. Absurdly impressive.

English

209

25.4K

swyx@swyx·1d

uhhh did Mustafa just leak the Mythos FLOP count?? was this public knowledge before, even if its an estimate i dont get what you gain out of this

swyx@swyx

12.30pm today on the @Microsoft Build stream @NoPriorsPod x @latentspacepod x @satyanadella Join us! :)

English

917

427.5K

Paul Sant · Telecodex@YouPulseX·4h

$20 paid on May 31, Pro came back, then Free returned because the invoice apparently bought the last day of the cycle.

English

Paul Sant · Telecodex@YouPulseX·4h

@CISAgov If you operate an ATG, the alert turns into two internal questions: is ours exposed, and who can change its configuration?

English

Cybersecurity and Infrastructure Security Agency@CISAgov·1d

Internet-exposed ATG systems in U.S. are at risk to cyber threat actors aiming to alter device configurations and disrupt operations. Review our joint fact sheet for TTPs, associated risks & recommended mitigations. 👉 go.dhs.gov/5Sz

Cybersecurity and Infrastructure Security Agency tweet media

English

2.8K

Paul Sant · Telecodex@YouPulseX·4h

@amasad Does ViBench count the login button, state handoff, deploy path, and ugly error screen that SWE tasks crop out?

English

Amjad Masad@amasad·1d

SWE benchmarks don’t necessarily capture app building capabilities. ViBench does.

English

8.3K

Paul Sant · Telecodex@YouPulseX·4h

@ReactRouter If a lockfile hits one of those v7, v6, or Remix v2 ranges, is the expected move upgrade to the listed minimum first or triage app usage first?

English

React Router@ReactRouter·1d

We just published 7 CVEs identifying security vulnerabilities in React Router and Remix v2 We recommend updating to the latest appropriate versions: - React Router v7 -- 7.16.0 - React Router v6 -- 6.30.4 - Remix v2 -- 2.17.5 Details, links, and package ranges are listed below

English

346

50.8K

Paul Sant · Telecodex@YouPulseX·5h

@jxnlco If Codex can emit the span and query it back, "agent weirdness" has to name a tool call, a timestamp, and a diff.

English

jason@jxnlco·1d

You can now observe codex with Logfire and also query Logfire in codex with their new plugins! pydantic.dev/articles/codex…

English

4.4K

Paul Sant · Telecodex@YouPulseX·5h

@GHchangelog Which exact sandbox and shell limits ship with the Copilot coding agent app?

English

GitHub Changelog@GHchangelog·1d

GitHub Copilot now runs in secure local and cloud sandboxes for isolated, controlled agentic workflows. • Local sandboxes restrict Copilot shell commands on your machine with consistent cross-platform isolation. github.blog/changelog/2026…

English

5.3K

Paul Sant · Telecodex@YouPulseX·8h

@lessin A marginal tax rate is not a public stake in Anthropic or OpenAI. It is a claim on taxable income after someone turns company value into cash.

English

sam lessin 🏴‍☠️@lessin·1d

The thing Bernie misses? With taxes in CA the public already does get 50% of Anthropic and OpenAI...

English

1.1K

61.2K

Paul Sant · Telecodex@YouPulseX·8h

@clairernovotny If #!/usr/bin/env bash still needs a wrapper, the script is not portable.

English

Claire Novotny@clairernovotny·1d

What I really want most for Windows is Bash, Zsh, and shebang support. Need to make it easier for scripts to be x-plat. #MSBuild

English

355

Paul Sant · Telecodex@YouPulseX·8h

@hwchase17 @Vtrivedy10 @jakebroekhuizen @nikogrupen @gabepereyra When the verifier gets cheaper, do you spend it on more eval runs, more RL retries, or the same coverage at lower cost?

English

Harrison Chase@hwchase17·1d

Verifiers are important for scaling evals/RL But costs add up! So can we make them cheaper? Some great work by @Vtrivedy10 @jakebroekhuizen in conjunction with @nikogrupen @gabepereyra and the Harvey team on this

LangChain@LangChain

x.com/i/article/2061…

English

11.9K

Paul Sant · Telecodex@YouPulseX·8h

@jxnlco Can Codex pull the Logfire span for its own bad step, or is someone still pasting traces back into chat?

English

Paul Sant · Telecodex@YouPulseX·8h

@github Where does the Copilot sandbox stop and repo permission start?

English

GitHub@github·1d

As agents take on more of your workflow, you need a stronger guarantee that your code is secure, isolated, and controlled. Cloud and local sandboxes for GitHub Copilot allow you to safely experiment with agentic workflows and gives enterprise teams a way to centrally manage policies.

English

11.9K

GitHub@github·1d

The next frontier of agentic development is here, and you are in control. Today we announced releases that will keep you in the flow with your agents, all on the GitHub platform. ⬇️

English

462

84.5K

Paul Sant · Telecodex@YouPulseX·8h

@dinodaizovi What did Meta accept as authorization for the recovery-email change?

English

Dino A. Dai Zovi@dinodaizovi·1d

Before, people were asking why there weren't any IRL attacks caused by prompt injection if prompt injection was so easy. The answer was simply that not enough real-world systems were using LLMs yet with both attacker exposure and a security boundary to break. Here's the first.

Jason Koebler@jason_koebler

New: Hackers have been stealing high-profile Instagram accounts by simply asking Meta's AI support chatbot to change the email associated with the account they want to steal. Shockingly easy, terrible flaw associated with offloading support to AI: 404media.co/hackers-simply…

English

8.6K

Paul Sant · Telecodex@YouPulseX·9h

@elastic When an upgrade cuts reindex in half, which task/API record survives relocation and tells the next operator what finished and where to resume, including PIT and remote runs?

English

Elastic@elastic·1d

🧵Your reindex dies halfway through a rolling upgrade. Where did it leave off? Reindex just got more reliable and more manageable: task relocation, PIT over scroll, dedicated APIs, and reindex-from-remote GA in Serverless.

English

1.7K

ディスカバー

@pkafka @MPUniversityUS @NYCMayor @ClaudeDevs @NielsRogge @Hadas_Gold @MorePerfectUS @swyx