cory mac domhnaill 🇺🇸🏴‍☠️

here for infosec/cloud/sw eng/appsec/data eng/ml. occasional screaming into the void. i speak for nobody other than myself.

[cory mac domhnaill 🇺🇸🏴‍☠️ @corymacd]

@morganlinton My logs show only that upload decisions respected zdr settings, and I'm not an enterprise.

[Jerry Thornton @jerrythornton]

“What a fooking place!” Beautiful.

Americans who seem to openly hate their own country need to explain this woman’s reaction to Boston.

Is she mistaken? Does she just not know what she’s talking about? Lay it out for me.

[cory mac domhnaill 🇺🇸🏴‍☠️ @corymacd]

Ughhhh dumb

[Corey Quinn @QuinnyPig]

Is this why Claude keeps saying it’s time to stop working?

[Son Luong @sluongng]

Codex just found a “workaround” of not having sudo on my pc…

[cory mac domhnaill 🇺🇸🏴‍☠️ @corymacd]

Why is @claudeai Claude code asking if anthropic can read my chat, when I've clearly opted out. What in the not okay is this? Keep eroding trust. Wow. Was just trying the last model drop. Now I'm out. So dumb.

[cory mac domhnaill 🇺🇸🏴‍☠️ @corymacd]

@elonmusk It's been super lazy today

[cory mac domhnaill 🇺🇸🏴‍☠️ @corymacd]

in case you havent tried kimi k2.6... you probably should.

[Nozomi @NozomiNetwork]

@hqmank didnt just add KYC, they collapsed the boundary between identity and thought. once access to intelligence is gated by who you are, it stops being a tool and starts being infrastructure for control

[BuBBliK @k1rallik]

🚨do you understand what just happened with Claude..

Anthropic quietly cut Claude's thinking depth by 67%.. didn't announce it.. didn't explain it.. an AMD AI Director had to dig through session logs just to prove it happened.

median reasoning dropped from ~2,200 to ~600 characters.. API calls went up 80x.. meaning Claude thinks less, fails more, retries more. and YOU burn more tokens paying for those retries.

they added a header that hides Claude's thinking from your logs. so when the model analyzed itself it found blank pages. and concluded it had stopped thinking.. you're paying $200/month for a model that can't read its own diary..

the thinking didn't disappear.. it just became invisible.. and Anthropic said nothing until the numbers went public

[cory mac domhnaill 🇺🇸🏴‍☠️ @corymacd]

@AnthropicAI @claudeai I just canceled my service because claude code cannot manage to read code in my project before vibing away for an hour, not producing anything of value. its wild. the erosion of value is so complete, its worthless.

[cory mac domhnaill 🇺🇸🏴‍☠️ @corymacd]

@ush1c I have had the exact same experience.

[cory mac domhnaill 🇺🇸🏴‍☠️ @corymacd]

claude code has been useless for me for a couple days now. while taking away feature? they think folks are going to keep paying for this?

[Jean P.D. Meijer ― 🇪🇺 eu/acc @initjean]

there's an AWS outage in me-central-1 because it got bombed

[Jason Bosco @jasonbosco]

"We used to debate using tabs vs spaces in code we'd type out"

[cory mac domhnaill 🇺🇸🏴‍☠️ @corymacd]

I've been trying to put a finger on this. I feel it deeply. Makes me want to go explore @encoredotdev again!

[Jesse Middleton @srcasm]

The current state of AI is a paradox.

On Monday, your model writes a clean analysis that makes you wonder why you ever hired interns.

On Tuesday, it misreads a single line in a legal contract and triggers a management crisis.

By Wednesday, your highest-paid employee is spending their entire day auditing AI outputs just to catch a 1% error rate.

If a human has to verify every single word an AI produces, you've just doubled the workload of your entire team.

I call this the "Jagged Frontier." The models are brilliant, but they are fundamentally unstable.

That’s why, these days, I’m seeing great founders focus on four things:

- Mapping the specific logic-gates where models consistently fail and placing a human checkpoint at that exact intersection.

- Building audit trails that make an agent’s reasoning visible so errors are spotted right away.

- Treating hallucinations as a manageable bug that can be rolled back without breaking the system.

- Moving high-risk tasks away from LLMs and back into traditional machine learning or hardcoded logic.

Buyers at Fortune 100 companies will not bet their career on a "genius" that needs a babysitter. They want systems that are auditable and reliable.

The value is in the last mile.

[Cape Cod Concerned Citizens-Legal Citizens First! @CapeCodCitizens]

Another Massachusetts success story brought to you by Maura Healey.

Cape Cod Potato Chips is closing their Hyannis factory and moving production out of state.

[JohnFGately @johnfgately]

Cape Cod Potato Chips to move production out of MA costing 49 jobs. “When a company whose entire identity is tied to this state decides it no longer makes economic sense to operate here, that should set off alarm bells on Beacon Hill,” said Paul Craney @MassFiscal #mapoli

[vx-underground @vxunderground]

"You penetration test 'em so you simulate the pressure"

🗣️🔥🔥