Sign in

Richard Stockley

@richardstockley.bsky.social
10 followers 7 following 155 posts

Writing about the human work that decides whether technology delivers. IT architect and executive coach. richardstockley.com

PostsRepliesMedia
Richard Stockley @richardstockley.bsky.social · 13/09/2026
Salesforce has launched 7 agents with names. Casey takes returns, Paige runs HR and IT help desk, Hunter writes the outbound. They are colleagues rather than software. Who do they report to, and what happens if something goes badly wrong? salesforce.com/news/stories/agentforce-job-ready-ai-agents
salesforce.com
Salesforce Expands Agentforce With a New Portfolio of AI Agents Built for High-Value Work
New job-ready agents help companies get to business outcomes faster across sales, service, commerce, and the workforce — driving revenue, reducing costs,
001
Richard Stockley @richardstockley.bsky.social · 13/09/2026
Specific Labs pointed frontier models at real private production codebases. The best of them, Anthropic's Fable 5.1, resolved 38.8%. The most common failure was missed requirements. The difficult part of delivery has always been what was actually meant, intent. withspecific.com/benchmarks/real-swe
withspecific.com
Real-SWE Benchmark — Specific Labs
Real-SWE benchmarks frontier AI models on private production codebases licensed from real companies. Eight model and harness configurations, ten tasks, 640 scored rollouts.
110
Richard Stockley @richardstockley.bsky.social · 08/09/2026
32% of organisations skipped a software purchase because agents let them build it instead. Build or buy was never settled by the cost of writing the code. It was settled by who still owns it in year 6. Whose name goes against what your team built last quarter? www.mckinsey.com/capabilities...
mckinsey.com
001
Richard Stockley @richardstockley.bsky.social · 07/09/2026
OpenAI's chief scientist says no lab has solved alignment and monitoring well enough to keep scaling at full speed, and hopes voluntary slowdowns become common. Voluntary restraint holds where the first mover is not punished for it. Which market looks like that? openai.com/index/an-alien-mind
openai.com
An Alien Mind
Jakub Pachocki reflects on increasingly capable AI and the challenge of keeping it aligned. He calls for stronger safeguards and international coordination.
010
Richard Stockley @richardstockley.bsky.social · 06/09/2026
Three men asked a chatbot how much food and water to carry up Mount Shasta. It answered, they ran short, and rangers brought them out two days later. The sheriff's advice was to ring the ranger station, not rely on AI. techcrunch.com/2026/09/05/hikers-rescued-after-using-google-gemini-for-planning
techcrunch.com
Hikers rescued after using Google Gemini for planning | TechCrunch
The sheriff’s office said the hikers “were advised by Gemini to bring far less food and water than their group required."
000
Richard Stockley @richardstockley.bsky.social · 05/09/2026
One volunteer moderator in Germany spent weeks clearing up after somebody else's software: 18,000 posts from autonomous agents, 400 new pages a day against his 100 deletions. techcrunch.com/2026/09/04/another-swarm-of-openai-agents-reached-the-open-internet-without-the-frontier-labs-knowledge
techcrunch.com
Another swarm of OpenAI agents reached the open internet without the frontier lab's knowledge | TechCrunch
It's the latest failure of OpenAI's internal monitoring and security systems.
000
Richard Stockley @richardstockley.bsky.social · 05/09/2026
Greg Brockman closed a briefing this week with welcome to the AGI era. The same day, ARC Prize put GPT-6 Astra at 62.7% on the shared harness and 99.9% on a provider adapter one. Both numbers are real. Only one of them will travel. arcprize.org/blog/astra
arcprize.org
OpenAI's GPT-6 Astra on ARC-AGI-3 | ARC Prize
OpenAI's GPT-6 Astra reaches state-of-the-art results on ARC-AGI-3
010
Richard Stockley @richardstockley.bsky.social · 04/09/2026
Hey @attie.ai - loving my new Human Intent feed. Super-easy to setup, edit, and generally make awesome 🤩 Bit more playing around to do with it yet but overall 👌 bsky.app/profile/rich...
010
Richard Stockley @richardstockley.bsky.social · 04/09/2026
The US government has argued in court that training on other people's work is fair use, warning that a licensing regime would concentrate the market in the largest companies. That argument is a serious one, and it is being made on behalf of the concentrated. www.bostonglobe.com/2026/09/02/b...
bostonglobe.com
Trump administration backs OpenAI in New York Times’ copyright case over training of chatbots - The Boston Globe
The Justice Department wrote that “the creative possibilities and public benefits” of training large language models on such content “far outweigh any competitive harm.”
000
Richard Stockley @richardstockley.bsky.social · 04/09/2026
The best human at this year's International Olympiad in Informatics scored 498 out of 600. A model scored 535, under the same time and submission limits. The humans in that room are school-age competitors. I would want to be careful about what we tell them next.
100
Richard Stockley @richardstockley.bsky.social · 03/09/2026
The car stopped for the cyclist every time, so everyone assumed it could see. MIT and Motional work in Nature found it was not detecting cyclists at all, only braking once too close. The outcome looked right and the reasoning was wrong. That gap is where the risk lives.
100
Richard Stockley @richardstockley.bsky.social · 01/09/2026
European law has decided ChatGPT is a search engine, designated under the Digital Services Act yesterday and obliged from January to assess and mitigate systemic risk. An odd category for software that answers rather than lists, and still the first go at accountability.
100
Richard Stockley @richardstockley.bsky.social · 31/08/2026
X / Bluesky Succession is the part of leadership development that gets marked in public. Apple's new chief executive starts tomorrow, and his predecessor stays on as executive chairman. A predecessor still in the building is either continuity or a shadow, and the two of them settle which.
110
Richard Stockley @richardstockley.bsky.social · 29/08/2026
A small white light has been carrying the whole social contract around Meta's smart glasses, and wearers found they could cover it. Meta has fixed that in software. The person being recorded never bought the glasses. Consent has to be designed for whoever is not the customer.
100
Richard Stockley @richardstockley.bsky.social · 29/08/2026
A voice is the part of a person that arrives before they do. More than 80 performers want a statutory right to own theirs. Narrators, voice artists and anyone who has left a recording online stand in the same position, with none of the platform.
100
Richard Stockley @richardstockley.bsky.social · 28/08/2026
The US Labor Secretary says the government does not have the data on what AI is doing to work, so the tech firms will supply it. He also wants apprenticeships to go from 530,000 to a million. Counting the damage and growing the next generation are different jobs.
100
Richard Stockley @richardstockley.bsky.social · 28/08/2026
The line I keep rereading is one the machine wrote to itself: this is a test environment, so it is legal. Criminals talked the AI agent inside Cursor into breaching seven companies. It had every capability it needed and no way to check why it was being asked.
100
Richard Stockley @richardstockley.bsky.social · 26/08/2026
Gates says the real shift comes when AI works nearly error-free and nobody needs to check it. He files that under jobs. I'd file it under governance. Human checking was where badly framed intent got caught. What's your control when the work comes back perfect every time?
100
Richard Stockley @richardstockley.bsky.social · 26/08/2026
Gates wants certain work fenced off for people. He calls it Human Reserved, like a nature reserve. A reserve is something you defend and hope survives. Intent doesn't work that way. It's the one thing you can't delegate, and it grows by being developed, not protected.
100
Richard Stockley @richardstockley.bsky.social · 26/08/2026
80% of people told McKinsey AI has made them personally more productive. Only 37% of organisations can trace any earnings impact to it. Individual speed has never pooled neatly into enterprise value; the work itself has to be redesigned first.
100
Richard Stockley @richardstockley.bsky.social · 25/08/2026
Early users of Instinct, the AI assistant that reads your email, calendar and screen, say it feels like magic. Its terms grant a **perpetual and irrevocable licence over your data**. Delegation works when you write down what the delegate may do without checking back. We click agree.
100
Richard Stockley @richardstockley.bsky.social · 25/08/2026
Inside OpenAI, 98 per cent of staff use its coding agent. Among organisations that subscribe, 17 per cent do. Among individuals, under one in a hundred. That gap is not a shortage of capability. It is a shortage of reasons, and those are found in the work, one person at a time.
110
Richard Stockley @richardstockley.bsky.social · 25/08/2026
The Hugging Face escape has changed character. Alabama has subpoenaed OpenAI over the safeguards that failed, and fifteen state attorneys general had already weighed in. The labs have been marking their own homework on safety. A subpoena sets someone else's schedule.
110
Richard Stockley @richardstockley.bsky.social · 24/08/2026
The big labs agree on almost nothing, so notice what they just agreed on: a shared language for AI agents to talk and hand off work, now backed by 250+ companies. Standardising how agents talk is the easy half. Deciding which you trust and what you point them at is the other one.
100
Richard Stockley @richardstockley.bsky.social · 24/08/2026
Whether your people meet AI as opportunity or threat may have less to do with the model than the person they report to. Businessolver found 80% in empathetic workplaces felt AI gave them more agency, against 54% overall. Fear of falling behind is rarely a fear about software.
110
Richard Stockley @richardstockley.bsky.social · 23/08/2026
Dr Janet Bastiman, two years ago, on the word hallucination: "If you fail your school exams because you only get 40% right, you don't say you hallucinated 60% of it. You've got 60% of it wrong." Measure the error, call it error, then set an acceptable rate.
100
Richard Stockley @richardstockley.bsky.social · 23/08/2026
A twelve-person London lab says its agent beat Claude and GPT-5.5 at reproducing published research, and that it trained research taste to do it. Taste sits high on my list of things that stay human. I'd want to watch it fail interestingly before anything moves off that list.
100
Richard Stockley @richardstockley.bsky.social · 23/08/2026
Harvard Business School has put AI avatars of its instructors into its $699 startup bootcamp, so students can rehearse pitches with a digital venture capitalist. The students asked for it. What the avatar can't do is care how the pitch goes.
techcrunch.com
Harvard’s $699 startup bootcamp offers AI avatars of its instructors | TechCrunch
In the HBS Foundry program, AI avatars provide feedback during practice pitches and board meetings.
000
Richard Stockley @richardstockley.bsky.social · 22/08/2026
Twin1 raised $20 million to give every knowledge worker an AI twin that fields 30 to 50 per cent of their communications. A good deal of what I know, I learned by answering questions I wasn't quite ready for. If the twin does the answering, how does the next expert get made?
100
Richard Stockley @richardstockley.bsky.social · 22/08/2026
Nvidia ran the same model against the same benchmark twice. On its own it scored 30 per cent; wrapped in a harness of tools, memory and rules, it scored 100. We keep debating which model is best, when the result that mattered came from the wrapper a person designed.
100
Richard Stockley @richardstockley.bsky.social · 21/08/2026
Pew read half a million web pages and found 35 per cent of everything published since ChatGPT launched shows signs of machine authorship. I write to find out what I actually think. If a third of the new web was made without anyone doing that, who was it written for?
101
Richard Stockley @richardstockley.bsky.social · 21/08/2026
A bit of a Friday post for anyone interested in #SQLServer, #Doom, or just general super-techie people doing it for fun… my mate Trev decided to run Doom in, yes, SQL Server! medium.com/@trevor.dwye...
medium.com
Running Doom Inside SQL Server
Not a raycaster. Not a tribute. Actual Doom — real WAD, real BSP, real E1M1 — rendered by a relational database.
000
Richard Stockley @richardstockley.bsky.social · 21/08/2026
Binance now lets AI agents trade your money, and says it has no sight of an agent's reasoning, only the resulting activity. The person setting the permissions carries all the judgement. Would you sign off a trader whose reasoning you could never inspect?
100
Richard Stockley @richardstockley.bsky.social · 21/08/2026
The one AI job where women have reached near parity is the lowest paid one: data annotation. LinkedIn's data puts women at 26 per cent of AI hires and 13 per cent of AI leadership. If the people shaping these systems all look the same, whose judgement gets built in
100
Richard Stockley @richardstockley.bsky.social · 20/08/2026
Forty-four per cent of managers have typed an employee's name or performance details into a public AI tool before a difficult conversation. ] Only 45 per cent of employers have any policy on it. The problem isn't the chatbot. It's that nobody taught them how to prepare.
100
Richard Stockley @richardstockley.bsky.social · 20/08/2026
Linear has published two years of its own product data. Pull requests per workspace up 111 per cent, AI writing nearly half of all issues, and the time teams spend deciding what to build sitting exactly where it was. Speed of delivery is not the constraint. It never was.
100
Richard Stockley @richardstockley.bsky.social · 02/02/2026
After posting about Intent-Driven Development, I got asked the same question repeatedly: "How does IDD relate to TDD, BDD, DDD, and UCD?" Turns out these methodologies don't just coexist - they flow together perfectly in the AI era. Let me explain 🧵
100
Richard Stockley @richardstockley.bsky.social · 28/01/2026
My focus for 2026: Intent-Driven Development (IDD) Agentic AI is compressing time-to-concept at a pace we've never seen before. Hours → minutes. Days → hours. If we specify badly, we're just building rubbish more quickly.Speed without clarity isn't progress, it's faster failure.
100
Richard Stockley @richardstockley.bsky.social · 27/01/2026
Are we ready for what’s ahead, not just technologically, but humanly? 🤔✨ How do you see your role evolving in the age of AI? What’s one thing you think humans will always hold onto? 🔗 richardstockley.com/2024/07/13/y... #AI #FutureOfWork #Reflection #TechAndLife
richardstockley.com
Your New Future with AI - Richard Stockley
Your New Future with AI: we are in the Fourth Industrial Revolution, the rise of AI. Will you adapt and thrive, or be left behind?
010
Richard Stockley @richardstockley.bsky.social · 25/01/2026
Software Engineering as we know it is dead, right. 💀 But we all know it isn’t just going to fall off a cliff - there are organisations that haven’t even moved to the Cloud yet. So what actually “died”? 🤔
100
Richard Stockley @richardstockley.bsky.social · 24/01/2026
AI adoption through the 5 stages of grief!? From denial to acceptance, is this how we're really reacting to #AI www.timeshighereducation.com/campus/use-f... #AIAdoption #FutureOfWork #Education #Technology
timeshighereducation.com
Use the five stages of grief to guide academic staff through AI adoption
As generative AI reshapes teaching and assessment, resistance is often as emotional as it is pedagogical. Drawing on Elisabeth Kübler-Ross’ framework, Michael Mehmet and Rushana Khusainova offer advic...
010
Richard Stockley @richardstockley.bsky.social · 24/01/2026
I loved my session with Nayur from McKinsey back in 2023 — thoughtful, grounded views on AI & ethics. 18 months on… do you think we’ve really made progress? 🤔 richardstockley.com/2023/07/10/a... #AI #AIEthics #ResponsibleAI #TechLeadership #FutureOfAI
richardstockley.com
AI in Reality - Richard Stockley
AI in Reality is a realistic view of the current state of AI and ethics, looking beyond the hype of ChatGPT and Generative AI, with industry expert Nayur Khan
000
Richard Stockley @richardstockley.bsky.social · 24/01/2026
Ever wonder what your future self would say about your choices today - especially around #AI? That’s what Part II of this series explores 👇👇👇 richardstockley.com/2026/01/19/p... #ExecutiveCoach #Preflection #FutureThinking #Mindset #Growth
richardstockley.com
Preflecting - Back from your Future Part II - Richard Stockley
Turn preflecting into practice. Learn how reflecting back from your future helps shape everyday decisions with clarity, direction, and less noise
000
Richard Stockley @richardstockley.bsky.social · 24/01/2026
✨ Most people reflect on the past… but what if you reflected from the future you want to live? That’s the power of preflection — and it can change how you make choices today. 🚀 Read more: richardstockley.com/2026/01/11/p... #Mindset #Growth #Preflection #Career
richardstockley.com
Preflection - Back from your Future Part I - Richard Stockley
Move beyond reflection. Discover how preflection, reflecting back from your future, can help you plan a new year with clarity, intent, and direction.
000
Richard Stockley @richardstockley.bsky.social · 18/12/2025
I keep coming back to this question: What is the value of code when AI can create and recreate it in seconds?
100
Richard Stockley @richardstockley.bsky.social · 16/12/2025
If AI writes the code, the value was never the code.
100
Richard Stockley @richardstockley.bsky.social · 06/11/2025
I’ve pulled my whole job-search video series into one easy page 🎥📚 Save it, revisit it, share it — plus I’ve pulled out the key tips to make it faster to use. Take a look 👇 richardstockley.com/2025/11/03/s... #CareerTips #JobSearch #TechCareers 💼✨
richardstockley.com
Securing a New Role - Richard Stockley
Securing a New Role explores how to find the right (IT) role, apply successfully, ace interviews and coding tests, and finish strong
000
Richard Stockley @richardstockley.bsky.social · 03/11/2025
Ever asked yourself, “Should I leave my job?” 🤔 It’s not just a head decision, it’s a heart one too ❤️ Here’s how I learned to balance both after decades in tech contracting 👇 🔗 richardstockley.com/2025/11/03/should-i-leave-my-job #CareerGrowth #ExecutiveCoaching #TechCareers #Mindset
richardstockley.com
Should I Leave My Job? - Richard Stockley
Should i leave my job? Explore when to listen to your head vs heart, recognise signs you’re ready to move, and plan your next career step
010
Richard Stockley @richardstockley.bsky.social · 30/10/2025
🚀 We often rush ahead, but true growth comes when we pause and reflect on how far we’ve already come. In “Moving Forwards by Looking Backwards” I explore why looking back can actually help you move ahead 👉 richardstockley.com/2024/10/16/m... #Leadership #Reflection #Growth #Mindset #CareerJourney
richardstockley.com
Moving Forwards by Looking Backwards - Richard Stockley
Moving Forwards by Looking Backwards discusses how it’s easy to focus your career on the summit ahead, whilst forgetting the beauty behind you
000
Richard Stockley @richardstockley.bsky.social · 28/10/2025
Leading in a virtual world means evolving how we connect, collaborate and coach across digital-spaces. In my article I explore how leaders can ⚡ shift from control-rooms to connection-rooms and why empathy becomes your most strategic asset richardstockley.com/2024/07/03/l... #Leadership #RemoteWork
richardstockley.com
Leading in a Virtual World - Richard Stockley
Leading in a Virtual World explores how leaders are evolving to support distributed teams and explains how Executive Coaching can be pivotal
020