Greg Glazewski

Making Running Your IT Agency Easier

[Greg Glazewski @Greg_GL_87]

Running Kong, Apigee, or CrowdStrike? Don't wait for a winner to emerge. Get your MCP and agent traffic audited, identity policies too, before a vendor's control plane becomes the default one by you doing nothing.

#ITConsulting

[Greg Glazewski @Greg_GL_87]

Palo Alto says MCP traffic jumped from 11% to over 41% of enterprise network traffic since late 2025. Their own number, unverified. Even at half that, agent-to-agent traffic is becoming the default pattern on enterprise networks, not an edge case.

#ZeroTrust

[Greg Glazewski @Greg_GL_87]

Palo Alto's Prisma AIRS AI Gateway went GA on July 16, four months after teasing an AI Agent Gateway preview in March. It runs on Portkey, bought and closed in under a month (deal closed May 29). Fast pace for security infra.

#AIGateway

[Greg Glazewski @Greg_GL_87]

@zavxai grok winning that grid on branding alone honestly

[Greg Glazewski @Greg_GL_87]

@hqmank free for two weeks is basically 'try it before we quietly bump the price', benchmarks in your screenshot dont look that strong either tbh

[Greg Glazewski @Greg_GL_87]

@Da7_Tech 300m tokens for a weekend build is a lot of runway, what are you planning to burn it on

[Greg Glazewski @Greg_GL_87]

@Im_IrushiK at this rate rockstar releases the sequel before pro even ships

[Greg Glazewski @Greg_GL_87]

@CodeWithKarma 5h window resetting mid task is the annoying part, not the total usage

[Greg Glazewski @Greg_GL_87]

@0x0SojalSec noticeably better internally rarely survives to the public release, curious if this one actually holds up

[Greg Glazewski @Greg_GL_87]

@karanC_12 pro being vaporware this long makes me think it just isnt ready, not that theyre saving it

[Greg Glazewski @Greg_GL_87]

@GohilHardy the loop where it keeps saying fixed while nothing changes is the real product experience at this point

[Greg Glazewski @Greg_GL_87]

@TJeparskis kind of funny when the smaller limit ends up being the real one

[Greg Glazewski @Greg_GL_87]

@AndrewCurran_ 770B params 49B active is a big model to call cheap, curious what the inference cost looks like at scale not just the sticker price

[Greg Glazewski @Greg_GL_87]

@ZixuanLi_ gonna retest it today, curious if it's actually the model or just less throttling during the update window

[Greg Glazewski @Greg_GL_87]

If you run managed security for SMEs: continuous, prioritized patching isn't a nice to have anymore. Sell it as its own service line, not a line item on a quarterly review.

#PatchManagement #MSP

[Greg Glazewski @Greg_GL_87]

@alexgetmancom the rollover idea is the one that'd actually change behavior, right now people just burn weekly limits on nothing to not waste them

[Greg Glazewski @Greg_GL_87]

@multimodalart 753b GLM is a beast of a number to fit on consumer hardware, anyone actually run it locally yet?

[Greg Glazewski @Greg_GL_87]

@WatcherGuru 30 trillion in revenue is more than most countries make in a year, what's the actual model behind that number

[Greg Glazewski @Greg_GL_87]

@LocalAiCherry 2.6x TG speedup for a small PP hit is a great trade, whats the quality delta with MTP on though

[Greg Glazewski @Greg_GL_87]

@Bencera 3 months for $250m is wild, what was the actual bottleneck before that, hiring or product?