Philip Taphouse

Founder of dotslash | Creator of the Service Operations Framework | Helping enterprises move from firefighting to autonomy | AIOps, Observability & Service Ops

[Philip Taphouse @philtaphouse]

Most enterprises are paying a £4.5M "Sludge Tax" every year. They aren’t paying it to HMRC. They’re paying it to their own messy, uncurated telemetry. If your team is drowning in data but starved for insights, read this. 🧵

[Philip Taphouse @philtaphouse]

@appleinsider I've had my air since the day it was released - love it!! I don't game or really care about the camera tbh - I love the beauty of the device - will be sad to see it go when I get the iPhone fold 😀

[Philip Taphouse @philtaphouse]

@thisdogcanbite @AppleTV yeah, I'll check that out.

[Philip Taphouse @philtaphouse]

@PauloJGTZ @AppleTV no, whats that?

[Philip Taphouse @philtaphouse]

@Aryan_Ahmed_ @AppleTV right!

[Philip Taphouse @philtaphouse]

@nickytrillgang @AppleTV haha

[Philip Taphouse @philtaphouse]

@AnxiousHolly 100%

[Philip Taphouse @philtaphouse]

@9to5mac @apollozac Nope!

[BigPanda @bigpanda]

AI is transforming SRE from reactive to proactive.

New Gartner® Innovation Insight shows how I&O leaders can eliminate toil, implement predictive insights, and achieve resilience by design.

Get the report: https://t.co/ONlI4K8gSk

#BigPanda #AIOps #SRE

[Demis Hassabis @demishassabis]

https://t.co/PTeDiv1b6L

[Philip Taphouse @philtaphouse]

Just floating around in the J-Space 😉

[ACTI @openacti1]

Acti is live! The Agentic Keyboard is here 🎉

🎁 RT + "Acti Agentic Keyboard" → Lifetime Premium
💰 Build a Skill → Win from $10K
🔨 Download & hold a key → See it yourself

📲 iOS & Android: https://t.co/pwjIswA6ll

Full story from @callmeyoung93

[Claude @claudeai]

Introducing Claude Science, a new app designed with every stage of research in mind.

Artifacts traced to their code, environments managed on demand, and 60+ optional scientific databases that you can connect.

Available now in beta.

[Philip Taphouse @philtaphouse]

Tomorrow dotslash at the home of the champions (I will never get tired of saying that!!), @Arsenal's @emirates stadium, as gold sponsors of @Dynatrace's #Innovate UK&I event.

We'll be talking about the Service Operations Framework (#SOF) and why #AIOps and #Observability only move the needle when they run as an integrated toolchain, not a pile of disconnected point tools.

Attending? Come find us at the booth.

[Philip Taphouse @philtaphouse]

Note to self… Try and stop thinking about how AI is changing the world and watch the @F1, just for a few hours!! Let’s go @LewisHamilton, @LandoNorris & @GeorgeRussell63!!!

[Philip Taphouse @philtaphouse]

@AndrewCurran_ This is a MAJOR turning point, I’ve covered this off in my Blog here (written from a UK perspective, but the argument still holds): https://t.co/CB5xAViOkl

[Philip Taphouse @philtaphouse]

If you're like me, you've spent the last few weeks watching half of @LinkedIn and @X post about "loop engineering" (yes I know, I need to get out more!). So what actually is it?

Well, a normal #AI chat answers once and stops. You ask, it replies, done. A loop is different - the model takes an action, looks at the result, decides what to do next, and repeats. It keeps going until the job is finished or something stops it. That cycle is what turns a chatbot into an agent, and it has been powering tools like Anthropic's #ClaudeCode & @OpenAI's #Codex under the hood.

Simply, loop engineering is the idea that the valuable skill has moved up a level. Instead of crafting one perfect prompt and watching the output, you build the harness that runs the model again and again on its own. You set the goal, let it work, have something check the result, feed the errors back in, and let it iterate until it's done.

@bcherny, who runs #ClaudeCode at @AnthropicAI, was recently quoted saying: "I don't prompt Claude anymore. My job is to write loops."

This gets powerful when a task is two things at once: repetitive and checkable. Repetitive, so automating it pays off. Checkable, so the machine can verify its own work without you reading every line. That second one is really (really!) important. No reliable check, no safe loop!!

Now the part that's being glossed over in all the excitement. An unattended loop with no checker ships bugs with total confidence, and the costs mount up quicker than you think. (One cautionary tale doing the rounds is the engineer who left a loop running overnight and woke up to a four-figure bill and a pile of broken code - I mention no names!)

If you're thinking about implementing loops, there are three golden rules to help keep you safe and avoid eye-watering bills:

1. A separate verifier that checks the work.
2. A hard budget cap (a token budget).
3. A clear rule for when to stop.

My honest take is that most teams don't need to rush at this, instead focus on getting your core workflow solid first. This is no doubt a trend that is here to stay, but without solid foundations, practices and standards around how you deploy agents and loops, it could lead to worse quality code at a higher cost.

I would love to hear your thoughts on this! Anyone out there implementing loops in their enterprise?

[Philip Taphouse @philtaphouse]

@SciTechera @Dr_Singularity

[Dynatrace @Dynatrace]

Strong ecosystems are built through shared success.

At Dynatrace, that means elevating partner impact and investing in what we build together.

📣 Dynatrace Partners: Join us for #Amplify2026 to hear directly from Dynatrace leadership and technology partners.

[Philip Taphouse @philtaphouse]

@joshdotai are you available in the UK 🇬🇧 yet?