jbkuru

Platform Engineer | On-device AI architecture & mobile security | AI capabilities vs. deployment reality | Peasant – Stay humble

[jbkuru @jbkuru]

https://t.co/NvqfrWi8QQ

[𝕏erias @xerias_x]

A true 9/11 survival story, as told by Brian Clark and Stanley Praimnath.

Two strangers, three floors apart in the South Tower.

Told as closely as I could to their own public accounts.

[Sterling Snow @sterlingmsnow]

I reread this every year.

I cry like a baby every year.

Todd Beamer died a father, husband, and hero.

Todd: Hello… Operator… listen to me. I can’t speak very loud. This is an emergency. I’m a passenger on a United flight to San Francisco. Our plane has been hijacked.

Lisa: I understand. Can they see you?

Todd: No. There are three that we know of. They have knives — razor knives, like box cutters. Someone announced from the cockpit there was a bomb. It sounded fake.

Lisa: Your name?Todd: Todd Beamer. United Flight 93.Todd: They killed one passenger in first class. They forced most of us back. Fourteen of us here. Five flight attendants. The guy with the bomb ordered us to sit on the floor.

Lisa: Are you okay?

Todd: We’re going down… wait. No. We’re leveling off. We changed directions. We’re flying east again.

Todd: A guy named Jeremy called his wife. She told him two planes hit the World Trade Center. Lisa, is that true?

Lisa: I have to tell you the truth. It’s very bad. Both towers are gone. A third plane hit the Pentagon. Our country is under attack. I’m afraid your plane may be part of their plan.

Todd: Oh God. Lisa, will you do something for me? Call my wife and my kids. Promise me you’ll call.

Lisa: I promise.

Todd: Our home number is… You have the same name as my wife. Lisa. We’ve been married ten years. She’s pregnant with our third child. Tell her I love her. I’ll always love her. We have two boys — David, he’s 3, and Andrew, he’s 1. Tell them their daddy loves them and he is so proud of them. The baby is due January 12th. I saw an ultrasound. We still don’t know if it’s a girl or a boy.

Lisa: I’ll tell them. I promise, Todd.(Lisa patches in the FBI.)

Agent: Todd, your plane is on a course for Washington. Best guess is the White House or the Capitol.

Todd: I understand. I’ll be back.

Todd: Everyone knows this isn’t a normal hijacking. We have decided we will not be pawns in their plot.

Lisa: What are you going to do?

Todd: Four of us are going to rush the one with the bomb. Then the cockpit. A stewardess is getting boiling water. We’ll take them out.

Todd: Would you pray with me?They pray the Lord’s Prayer. Then: Yea, though I walk through the valley of the shadow of death, I will fear no evil, for thou art with me.

Todd: God help me. Jesus help me.Are you guys ready? Let’s Roll.

[jbkuru @jbkuru]

@tobi Yeah. Teams forget the actual job. They keep stacking the utopia while customers are out here screaming for help. Blinded by busywork in their own bubble. It’s a balloon. One simple solution pops the whole thing.

[jbkuru @jbkuru]

@SledgeDev

[jbkuru @jbkuru]

@burkov It is like punch cards to Assembly phase !

[jbkuru @jbkuru]

@defyneric @XMoney X money requires VPN to be turned off is that so ? Why ?

[Cloudflare @Cloudflare]

At 250 billion DNS cache entries, one wasted byte costs 250 GB of RAM.

Five Rust optimizations later: 100 TB freed, inserts 43% faster, lookups 19% faster.

We didn't trade speed for space. https://t.co/mMyOYnj0mQ

[jbkuru @jbkuru]

Most “VoIP is broken on mobile” tickets are not VoIP failures.
On a locked phone the SIP socket is already gone. What rings you is a push that tells the OS to build a call screen from zero. The actual voice path starts after you swipe.
https://t.co/tw4aYOeEZq

[Lara Logan @laralogan]

SHE BUILT THE TOOL DOGE USED TO FOLLOW THE MONEY
Before https://t.co/fCs1WbyY64, the public could search government awards by year or recipient, but not the actual language buried inside their descriptions.

Jennica Pounds changed that.

The deaf database engineer built a tool that allows anyone to search specific keywords, uncover questionable government spending and trace taxpayer dollars as they flow through the NGO industrial complex. DOGE used her work, and now the American people can use it too.

In Episode 95, Lara Logan sits down with the woman behind DataRepublican to examine the money networks involving USAID, activist organizations and influential figures such as George Soros. They also discuss Jennica’s remarkably accurate 2024 election analysis, the personal cost of being publicly exposed and how her Christian faith helped her endure it all. @DataRepublican | @DataInterpretr | @GoingRoguewLara

WATCH EPISODE 95:
https://t.co/dHyKgeiH11

SUPPORT OUR SHOW:
https://t.co/ZGF3YDTZzv

#LaraLogan #DataRepublican #DOGE #GovernmentSpending #FollowTheMoney

[jbkuru @jbkuru]

Dung's Argumentation Framework (1995)

The core idea: An argument is not just a claim. It is a claim plus a reason. Arguments can attack other arguments. The question is not "is this claim true?" but "does this argument survive all attacks against it?"

[jbkuru @jbkuru]

Build with Grok

[jbkuru @jbkuru]

Hexagonal Architecture for Non-Deterministic Compute

We dropped language models into systems built for deterministic code—and handed them write access to the database.

The model isn't the problem. It is exceptional at proposing, but terrible as a source of truth. When an LLM can write state, invent tools, or execute raw PDF instructions, system drift is guaranteed. That isn't a prompt engineering failure; it's a missing structural boundary.

I’m building an agent stack around a strict separation of concerns:

The model proposes. Code verifies. A deterministic layer decides if state may change.

Untrusted document in. Structured proposal out. Nothing touches the system of record unless deterministic code says yes.

It sits across three planes:

Control: Domain aggregates hold identity and hard rules. The model offers structured proposals; pure code enforces state mutation.

Isolation: Untrusted extractions run in an isolated sandbox, physically air-gapped from write credentials.

Evidence: Append-only execution trails ensure every run is auditable and replayable.

It's Hexagonal Architecture applied to non-deterministic compute.

Building it. More when it can fail a test suite, not just a diagram.

[jbkuru @jbkuru]

Wear helmet please

[jbkuru @jbkuru]

I was staring at the confidence labels on the graph again last week.

CONFIRMED. PROBABLE. INFERRED. Gaps marked RISK. Everything had a citation, which felt responsible, but the labels still sat there like polite guesses. A call chain could be “confirmed” with a Bitbucket line and a Confluence page and I still wouldn’t bet the farm on it when the downstream test turned red.

That bothered me more than it should have.

I kept thinking: we have the source, we have the claim, we even track when the claim first appeared. So why does it still feel like belief instead of something closer to knowledge? And what does the system actually know that it *doesn’t* know?

That’s when I fell into epistemic logic. Not the abstract academic version. Just the basic distinction between knowing something and believing it, and the idea that a justification can be treated as a first-class object instead of a footnote. Justification logic in particular clicked because we already store the citations. The missing piece was treating those citations as actual reasons that can be combined, weakened, or defeated when new evidence shows up.

I’m not rewriting the whole system around it. But the next time a gap gets classified or a chain gets promoted, I’m going to ask a slightly different question: is this something we believe with good reasons, or something we’re willing to treat as known for the decisions that depend on it?

Turns out the difference matters more than the floating confidence number ever did.

[jbkuru @jbkuru]

@svpino I feel guilty if I don’t understand the code so I end up reading all of it but the volume is just so much I’m never sure I actually get every part of it.

[jbkuru @jbkuru]

The shift in behavior between 1-on-1 chats and group chats comes down to audience dynamics, social signaling, and reduced accountability. When a conversation moves from a private exchange to a group setting, the psychological motivation changes from connection to performance.
In a 1-on-1 chat, the only person you need to impress or convince is the person you’re talking to. In a group chat, everyone else becomes a silent audience. Dropping provocative takes, “winning” an argument, or displaying edginess becomes a way to claim social standing, project authority, or define a role in the hierarchy.
This creates what you might call the spectator effect. A provocative comment is often a test — seeing who laughs, who seconds the opinion, or who reacts with a specific emoji. Knowing that silent observers are watching makes people feel bolder than they would alone.
Responsibility also gets diffused. In a private exchange, any tension or hostility lands directly on one person, making the conflict explicit and immediate. In a group, the cost feels shared. Accountability drops. It’s easier to treat the group as a stage rather than a set of individual relationships.
Text itself amplifies this. Without eye contact, tone of voice, or real-time rapport, empathy falls. People become more willing to use others as a sounding board for hot takes instead of treating them as people with feelings.
There’s also a low-risk escalation effect. You can drop a strong opinion into the group, get the dopamine of asserting it, and then step back. You avoid the real-time friction or awkward silence that would follow if you said the same thing face-to-face.
Here’s the contrast in simple terms:
1-on-1
• Primary goal: mutual understanding or direct coordination
• Empathy barrier: high
• Conflict risk: personal and immediate
• Accountability: direct
Group chat
• Primary goal: public signaling, status, or entertaining the group
• Empathy barrier: low
• Conflict risk: abstracted and performance-based
• Accountability: diffused
The most effective response when someone starts performing in a group is to starve the performance. Responding aggressively inside the group validates the stage they wanted. Ignoring the bait or moving the conversation private breaks the cycle.
The medium changes the incentives. Once you see that clearly, the behavior becomes a lot less surprising — and a lot easier to handle.

[jbkuru @jbkuru]

@brithume It looks awkward. My two cents: seeing how the IRGC moves elements on the board and then picks targets, instead of all-out war, makes tactical sense. Drip, drip, drip.

[jbkuru @jbkuru]

Chicken wings

[jbkuru @jbkuru]

@Milajoy Trapped in a bubble, validated only by their social circle and echo chambers.

[jbkuru @jbkuru]

@BrentBritt58901 Jesus help them