IT Manager & Strategic Advisor | Orchestrating Sovereign Intelligence for AUM Entities | Specialist in System Innovation & Operational Excellence | Le Mans Winn
Nederland
Joined May 2008
- Tweets27K
- Following5.9K
- Followers5.2K
- Likes15.8K
The future isn’t a wave to survive—it’s a course to steer. Stop waiting for the ultimate tech tool. Grab a single sheet of paper and answer three questions: Where do we lead? Where do we draw the line? What must our people learn? Clarity brings control. 🚀 #AI #Leadership #SME
1
17
Old TODOs now cost real money.
Google found that coding-agent sessions chasing stale TODOs used 77% more tokens and 105% more wait time. 41% chased bugs already closed.
Great news, it is cheap to fix: clean stale TODOs before your agents read them. teapaw.org/glenngabe/status/21190…
From a Google software engineer and data scientist. Your TODO comments could be distracting agents -> Agent Distraction
"As agents' agency through tools and skills grows the ability for side-quests scales proportionally. Obsolete specs, long resolved bugs, and lagging documentation no longer rest in long forgotten corners of the codebase, but rather are unintentionally ingested as context indirecting and misdirecting the next wave of coders - the agents. This indirection is not free: in the best case it only costs tokens, steps, and time - in the worst case it distracts and grounds the agent on falsehoods leading to incorrect logic and decisions during implementation."
And what they did about it:
"As a result of this multi-layer approach, our todo cleaner has been able to remove thousands of actually obsolete references over the last 6 months exceeding 95% acceptance from human reviewers."
developers.googleblog.com/ag…
2
16
Without shared meaning, AI keeps guessing. Franklin Templeton only got useful research agents after they built a semantic layer — what the data means and how it connects — before scaling. Define meaning first, then let the agents run. openai.com/index/franklin-te…
13
Your cold chain is an attack surface.
FBI and CISA warn of attacks on unpatched, internet-facing refrigeration controllers. European waves hit ~250-350 French stores, then over 115 stores in six countries.
Pull controllers offline. Change defaults. Treat OT as production.
18
OT blind spots are a board issue.
Palo Alto Networks surveyed over 1,600 critical infrastructure leaders: 68% lack complete real-time OT asset visibility, and 95% worry about frontier AI-powered attacks.
Map every connected OT asset first. Then automate where downtime hurts.
15
Patches at machine speed.
In one run, an OpenAI pipeline validated and patched 948 Linux kernel privilege escalation flaws, at about $1,279 each.
Defenders now have that speed too. Make sure your patch process can keep up. teapaw.org/OpenAI/status/21181194…
Our security researchers built an autonomous cyber vulnerability discovery, exploitation, and patching pipeline.
In a single run with an upcoming cyber model, it validated and generated patches for 948 local privilege escalation vulnerabilities in the Linux kernel at a total API cost of $1.2 million.
23
Sovereignty you can test.
On 24 October, AWS ran its European Sovereign Cloud for several hours with the global backbone switched off. Workloads kept running.
Proof beats promises. Ask your cloud provider when they last tested it, and ask for the report. aws.amazon.com/blogs/securit…
10
Your freezer is an endpoint.
In October attackers hit about 700 French supermarkets via refrigeration controllers left online. Now Belgium, Ireland and Sweden report the same.
Good news: the fix is known. Map every device suppliers reach remotely. Take it off the open internet.
10
Deloitte: 40% of companies let AI execute decisions without human approval, at least to a moderate degree. Only 15% report a significant shift in decision-making authority for their people. Give your people the room you give the machine. deloitte.com/us/en/about/pre…
15
Gerard Grouve retweeted
Dus t kan wel. Voortaan die klimaatwappies net zo aanpakken. telegraaf.nl/binnenland/live…
Gerard Grouve retweeted
Zo @Merelek vraagt even aan #Jetten of het nou verstandig is om mensen die #PVV gestemd hebben in de hoek te zetten.
Dat minachtend lachje van hem zegt meer dan z’n antwoord. #Persconferentie
From AI talk to AI results.
On Q3 earnings calls, 74% of S&P 500 firms mentioned AI, but only 5% reported a measured result in production.
That gap is great news for teams that measure. Pick one process, baseline it, and report the gain every quarter. teapaw.org/azeem/status/211603712…
Among S&P 500 firms holding Q3 earnings calls:
74% mentioned AI
18% claimed an impact
5% reported a measured result in production
AI’s next test is whether this gap closes.
1
1
60
Plan the work before the headcount.
Gartner predicts 30% of staff replaced by AI will be rehired by 2029, often at higher cost. Only 19% of firms plan their workforce around how AI changes work.
Start with the work your strategy needs, then pick the mix of people and AI.
23
Simpler agents win.
Anthropic's applied AI team now advises one agent, one loop, and subagents only when truly needed. One customer cut runtime in half by merging subagents.
Fewer moving parts also means fewer permissions to govern. Start small, measure, then grow.
1
1
37