This page is not created by, affiliated with, or supported by Slack Technologies, Inc.
How many others are building a bench to augment key problems? Anyone thinking about this in the AI acceleration space? :thinking_face:
First rule of Disney spending. Why buy one when you can have two for twice the price.
Feel like there is at least another 2 hour presentation from @jason.cox @alexi.s.varanko
@jason.cox can you share any more details around the "JEDI" program. Idlf memory serves, this is the program at Disney that helps train people to be effective Enablement engineers correct? I'm in an organization that is standing up an Enablement engineering group, and any insights/materials/subject matter would be helpful.
:heavy_equals_sign: Please welcome Steve Smith, Global VP of Technology at Scale, Equal Experts here to present 3 Ways You’re Screwing Up Platform Engineering - And How to Fix It
Yes, @mvk842 kindly facilitated a joke at my request. It was perfect and a great example of the lengths IT Rev goes to help their speakers have a great experience ♥️
Jason's case is a specific example of Don Reinertsen's insights in this blog - the problem is not with shared services per se, the problem is when they are measured on unit cost. "Curiously, whenever a shared resource has high capacity, consistently low latency, and fast response times, nobody wants the headache of managing a smaller, local resource to replace a satisfactorily performing shared resource" more: https://www.forrester.com/blogs/expertise-and-the-shared-services-problem-a-conversation-with-don-reinertsen/
Examples of Power Tools: Kubernetes, Kafka, Istio... In other words, things that you don't want to run on your laptop. Hahaha.
Also, tools that people are drawn towards using for their power, their utility, their capability, when a screwdriver will probably just do today's job
With platform Engineering, TCO (Total Cost of Ownership) is paramount. Why building in exit ramps is just as essential as onboarding any new technology.
I've not heard it described that way before. Martin Fowler said years ago something like: remember today's modern service is tomorrow's legacy, or something like that. Reversability is how I look at it, from a Kent Beck quote a couple of years ago
Let’s build the cluster of our dreams! (Most platform teams don’t say that out loud)
It's important to assume good intentions, I think a lot of people just get sucked into perfectionism and robustness, over good enough and resilience. There's some resume burnishing going on, I guess
“When do power tools happen? When you work inside out on what you want instead of outside in on what development teams need.”
Spent some time on that 😄, I wanted some memorable but not too critical. Glad you liked it Jeff
What a relief to see cost effective tools elevated above the “default” choices. I’m usually labeled an uncooperative renegade when I point out the simpler solution 😬
But...not just about the tools and preferences. Remember, these platforms and processes may contain all the "compliance as code" value. Now what? Distribute that 3x? = diff costs.
"Kubernetes is hard to get right, which means it's easy to get wrong" this is true of many power tools. It might be the right tool, but should not be signed up for without understanding the consequences and TCO behind it.
uncontrolled variation in base tech is a huge issue. Hearing more and more pain around it.
A colleague said it best: “if we’re going to customize, we need to have a long term maintenance plan”
When you give your teams optionality they take it, and find creative ways to create new foot guns...
"Each week, just ask what % of time is spent on unplanned tech work on Google Cloud Platform configuration, etc."
ShadowIT is what crops up when the barrier to entry and usability is too high or you're not meeting needs. Those are opportunities to identify feedback from your customers on what their needs are. Embrace ShadowIT instead of knocking it down. learn from it.
and when your environment is so controlled and locked down, that even shadow IT won't work...anger and no throughput.
that's a cultural problem. ShadowIT shifts to credit cards and personal devices....... Engineers will find a way. :rolling_on_the_floor_laughing:
Yes that's a part of it... bake in your technical alignment, and your tech stack will be a big part of that. But there's other aspects of contextual alignment you want to bake in e.g. SLO settings
Polyglot platforms is a matruity aspect. you can offer multiple language support, but cannot be a first mover.
Hmmm... interesting. Not sure I agree, it'd be good to talk about this, and hear more about your perspective (I'd say multiple tech stacks in your heritage services is inevitable in heritage, I wouldn't advocate it for modern workloads)
Let’s chat. We offer multiple languages today while offering the same rails experience. It’s allowed for more adoption by developers and increasing throughput and ttm
Would you and @hany.elemary like to duck out of a morning session, and we catch up outside Azure? Are there any particular GenAI sessions that are must-see for you? :thinking_face:
Oops. Apologies I'm just now seeing this, folks. I have better availability tomorrow, if that works.
Steve undersold himself. He was a huge help at SiriusXM getting our platform and Reliability teams up and running!
Tickets for automatable work is a public announcement that you're failing at operational agility.
Are tickets in any form a universal tech-disease predictor? I’ve never seen a ticketing system associated with a healthy environment.
ok, if they trigger an automated flow....and are really serving a history logging purpose.
Put the things you don't want people to do behind a ticketing system
I actually think of it the other way. Tickets for routine work should just be self service or automated: Run a job, provision your own service/VM, etc... Tickets for creative work(features, stories, etc.) are vital.
And the routine work ticketing problem is a continuum. You may need to enter tickets for everything to begin to just collect and understand the work. Then, remove those tickets with self-service automation
My point of view is that routine work should be automated or self serve. Features, etc., should not be initiated by ticket. If a feature needs to have some surveillance over its life cycle, the development process should be instrumented to provide that surveillance and the instrument should not look anything like a typical Jira ticket.
That may be a difference in terminology Scott. I don't consider your examples tickets in our workflows.
Ok. I guess the general term would be: WorkItem? And there are different classes of those: Service Request, Incident, Feature. The default is often to create a Service Request for everything and have folks process those on a queue. That is bad and should be automated. Creative work in a SDLC is managed with a WorkItem type of Feature..
Agree that tickets also are a key signal for poorly understood need. Routine work should be automated, but in some cases you can't get there yet. My frame also includes EUC and customer support. I think we also need to clarify some language. Is a fully automated ticket still a ticket?
I have seen fully automated tickets to support traceability. Pipeline opens change record, pipeline deploys, pipeline closes ticket.
I've also seen 17 tickets for a provisioning flow that were not automated. Ironically those 17 tickets(in Jira) were created from one Jira ticket and then assigned each of those 17 tickets to different teams.. Yes. That is a disease.
Why does anyone think you should build everything around the complaints department?
The decision to automate vs tolerate ticketing can be hard - how to prioritize? This is why platform managers must be product managers.
TPP - unplanned work is often a bottleneck and constraint. need to identify that workstream and protect the teams from it and unlocking more deep work.Interested in how others approach this today.
we utilize a patrol system to protect the team except for "the tribute" for that day/week. this allows 1-2 persons to be t-boned with unplanned work and protect the others on the team. rotation through this is important as it also fosters context sharing better
"trial by fire" for newcomers. diving in to problems is the fastest way to learn things.
Platform “products” work well only when platform teams do the same high-quality discovery that other successful product teams do.
Totally agree. I don't think platform product manager or platform user researcher are anything different from a product manager or user researcher in a platform domain
Definitely invite @steve.smith back for more cultural criticism...
✨Next up is Chuck Lafferty, Vice President - CRM, and Dr. Mary Hayes, Director of Research - People and Performance here to present The Levers Of Leadership at ADP
"we do everything from hiring to retiring" Nice. (Also rhymes: firing)
The definition of hybrid is simple. It is an employee who works sometimes in the office/ on-site and sometimes remote. We did not put any quantity of days in one place or another. So again we let the respondents choose between three options Office/ on-site, Remote only, and hybrid. Hope that helps.
1I’d imagine one person in the office and the rest nowhere near the office would have less engagement
"2.8x more likely to be 'all in'" "10x more likely to say 'I'm part of the best team'"
"At least 2x more likely to say they're on the best team when they have complete trust int heir leader"
“The only person who can influence the engagement of the team is the team leader.”
I'm gonna be a contrarian on this one. I've definitely been able to impact and inspire engagement and morale from within a team....even impacting the leader.
The items I shared yesterday were designed to be a leadership effectiveness tool to help leaders learn from their team if they are providing what each team member needs. But as we know in measurement these eight might not be the only parts of the story. One of the questions talks about teams having each others back. This is a important part of the leaders need to foster team work. I have had some ineffective leaders and some times the team goes it alone and a natural leader comes from within.
"The only person that can influence engagement is the team leader" — "so we decided to make the leaders responsible for engagement"
"Student gets 5 As, and 1 D. Which subject do you discuss first? Org psychologist says 'probably the As first'"
An old boss used to do this for team performance reviews. Focus on building on your strengths!
Use the bright spots to figure out what works for this human, use what already works to figure out remedies…
Looking at the positive also helps build self efficacy in an individual. Often times we always tear our people down by only focusing on the negative. Keep looking at the positive in those that follow you. It will go a long way to building trust.
Focus on what is working to translate a recovery from what is not working
"orgs are just a network of commitments" — so sharing loves vs. loathes enables a value exchange. Neat.
"it is not enough to measure, you must take some action" :thinking_face:
I see a TON of measurement efforts that never lead to action. It’s worse than not measuring at all. It callouses people to the whole deal.
Absolutely. Measuring and not closing the loop leads to survey fatigue and loss of trust.
Team Topologies!!!!! such a fantastic read and so many things to learn and implement.
And now, Manuel Pais, Co-author of Team Topologies, and Dr. Laura Weis
Head of Future of Work, Satalia/VML, here to present The Science Behind Team Cognitive Load and Why It Matters for Engaged Teams
DevOps topologies is a great infographic https://web.devopstopologies.com/#
Very interesting topics, is there an online resource that lists the key experiments and studies in these areas?
Thank you, that’s good background information and has a reading list for more investigation, which is the next step in looking for the key experiments. The article sometimes uses Team Cognitive Load in places where context makes clear overload is what is meant. Good stuff!
Whew. A talk on cognitive load and attention at the end of a long AMAZING day of sharing info. I am definitely going to rewatch this talk when I have more attention credits.
Very interesting talk, but yes, there is a bit of irony to this content being in the 10th hour of a conference! (if I'm mathing right, which I might not be at this point? 🤪)
Super curious how AI (read ML) is used to better understand the complex graph theory behind cognitive load. Feels like the data collection part of this is a big challenge!
ETLS/DOES is a wonderful experience of overstimulation and positive cognitive load for many days... 🙂
This is fascinating, I see cognitive load impact every day from teams, but the explanation on how this happens is Amazing, I feel like we have hope to help!!
Seen this load play out most when Sr. leaders struggle to prioritize and say no, and have no tools but "produce more!" Adding more, when the last thing is not done . #KanBanWIP
And how does this align with the findings of Cognitive Psychology in terms of the Beliefs held by the individuals experiencing negative emotion related to cognitive load -- or not
Awesome food for thoughts in the presentation from Dr Weis and Manuel Pais. Wow. I’ll skip cocktail 🙃
Who is here as part of the Team giving the feedback? Say hi
Love this end of the day review hosted by @genek... such a good idea for fast feedback and review.
It has been great to connect with so many on the Southwest team. They came curious!
Thank you to all speakers for such an informative and thoughtful and amazing day 1! 🙏 🙏 🙏
Friends - please join us at the industry party right after this!!!!
ETLS Las Vegas Livestream—Get access to days 2 & 3 for your colleagues here: http://itrevolution.com/membership • $2,500 for a 6 month IT Revolution Enterprise Membership with unlimited users • They can join this amazing Slack community and engage with the talks live—they'll even show up on the StageSYNC feed! • Plus access ETLS Connect virtual events (with in-person watch parties hosted by Equal Experts) starting on October 23
<!here> (the rare at-here message): If you have colleagues you couldn’t bring, ^^^ is a great option so they can watch all the talks with you tomorrow morning!!! Please let @alexb know if you have any problems: Thank you, all, and see you later tonight and tomorrow!!!! 🎉🎉🎉
If you left a swag bag in the afternoon plenary, you can now find it at the Registration desk.
Thanks everyone for the great feedback on my platform engineering talk today. Special thanks to @mvk842 @annp for allowing me to pick my own intro music and sabotage it 🙇 Some folks have contacted me with their own platform engineering screw up stories. Please do get in touch, always happy to listen and share 😄
Reminder: Get yourself to your seat in the Azure Ballroom for the opening remarks. We’re kicking off Day 2 in 15 minutes at 8:45am PDT! https://devopsenterprise.slack.com/files/UATE4LJ94/F07H7N10D62/timer.png
Great day 1!! Looking forward to see what unexpected things Day 2 brings 😹
ETLS Las Vegas Livestream—Get access to days 2 & 3 for your colleagues here: http://itrevolution.com/membership • $2,500 for 6 months of enterprise access with unlimited users • Everyone can join this amazing Slack community and engage with the talks live—they'll even show up on the StageSYNC feed! • Access ETLS Connect on October 23 DM me with any questions!
✨ Starting us off this morning, one of the teams from Vanguard, Devlin McConnell, Senior Manager - Emerging Technology, and Matt Butler, Director of Analytics and Automation - Center for Audit Practices and Enablement, here to present: Exploring What GenAI Can Do for Vanguard Crew and Clients
Pdf version of megatrends research: https://corporate.vanguard.com/content/dam/corp/research/pdf/megatrends_ai_demographics_and_the_us_economy.pdf
2Seeing Dr. Joe Davis talk was a seriously geek out moment for me — it was amazing hearing him take a minority view that AI can materially change GDP growth. You might see me in the audience in some of these shots. 😂
(Usually you'd see Dr. Joe Davis talk at conferences where the other speakers are Fed chairs, people who set economic policies, former US presidents, etc. It was an amazing talk, and I think should inform how tech leaders make strategic decisions around AI.)
To watch the other four talk Dr. Joe Davis talk excerpts, Vangaurd has posted them on YouTube: search for "Vanguard joe davis megatrends"
Key point: using AI WITH experts, not instead of.
Let's face it: Internal audit has a reputation of being conservative adopters of technology. To see internal audit exploring ways to use GenAI is... pretty freaking wild and awesome.
In regulated environments, the effectiveness and efficiency of your risk and audit functions can make or break your agility and focus.
Using GenAI to power interal audits scopes is bold! Love it! 👌
I think it's amazing to even imagine audit plans being actually designed around relevant risks, and scoping it accordingly. As much as I love the audit community, most audit plans look pretty close to "what we did last year."
Annual planning of ‘what to audit’ seems slow. If you’re using AI to help determine what to audit, would it make sense to iterate the audit plan more frequently?
Great point. Audit orgs still seem to be “plan and execute” orgs in a world increasingly demanding “sense and respond” behaviors. A continuous, machine-learning driven, fluid action plan for “what we be drawing scrutiny right now” would seem to be a true modernization of “audit.”
It seems like a natural progression if you’re looking to move from a defensive role/posture to an offensive one.
Yes! Try inviting audit to quarterly business planning. Audit can be an awesome partner in changing level 3 systems.
Auditors, audit others with AI.... Interesting conundrum on auditing with a tools that itself needs to constantly be audited. Very interesting and potentially a coup to audit turnarounds
I’m definitely seeing a trend of training AI on hyper focused internal language and translating that across domains and regions. Audits, Semantic Issues, etc
Much better than using it to write a song about my dogs 🐕
I think you might be onto something. Let’s make a slide deck 😂
Or just one giant embedded video and we can take the day off
Having the technical leadership (not just consultants) in the risk org will be crucial to make this pivot. Bringing in signals, systems thinking..etc.
Oooh, what an interesting culture shift! “People who usually play defense, learning to play offense!”
Using AI to generate an index over a body of content is one of my favorite use cases, and IMHO is one of the safest.
Curious - how are teams who have implemented GenAI measuring the gains? What are some of the results folks are seeing?
Follow on question: can traditional DORA metrics be used?
In regards to developer productivity gains we discuss that here. https://itrevolution.com/product/devops-enablement-with-genai/
3👟 And now, please welcome Fernando Cornago, Vice President Digital Tech, adidas here to present Journey to Composability in Commerce: Boosting Efficiency and Innovation with GenAI at adidas
Change lives through sports! Love it and doing it with IT Enablement!
“Did we just need to have muscles where it matters?”
Yes! #GetSmall more with less and smaller teams. Love Adidas 50% resource reduction with better results. You can do it too!
Had a team go from 50+ to 22 engineers (organically, via natural attrition), getting more done with higher quality. At first, as we didn’t replace people, people were nervous. It helped when we reframed it as being better. Doesn’t getting better mean that we should be able to do more per person? Isn’t that proof of being better?
Their retail channel partners (Dick's Sporting Goods, etc) now building great e-commerce, growing 10%+ annually. Now adidas focusing on wholesale part of business, less on directo to consumer.
It's anarchy 😄 I covered it in my talk yesterday https://www.equalexperts.com/blog/our-thinking/how-to-overcome-technology-anarchy-with-aligned-autonomy/
Adidas: first in China on TikTok, grew to 30% of revenue! Tried in Europe: totally failed: resulted in 4 sales.
I never imagined I would be hearing about Generative AI from Adidas. Amazing to see impact of technology!
I was just texting with him this morning! It’s not the same without MikDLT. 😢
Just ask the simple question; “What are the 3 things you’re going to do every quarter?”
“We used the Flow Framework to figure out what part of our spend we wanted to capitalize - build versus run” 💡
“Pure IDE” time is not a measure of engineering effectiveness. Just sayin.
I thought about this pure IDE time and the speaker did a good job with the graphic breaking down time used. Perhaps it would be better reframed as reducing ineffective time or toil
This would never happen, but I could be spending 10 hours a day working on the WRONG thing...
And “talking” to CoPilot counts as “pure IDE” time if you do it in the IDE. But talking to real live people doesn’t….
Thanks all for the comments, fully aware of developers not being measured by time on ide but on value. However, the exercise has been useful to confirm that the teams that transformed are much productive (and happier). But as said, the teams working in legacy and dependent systems are there because we chose not to invest there. Not to be fast there.
We are using the exercise now as a benchmark for teams that self-report they don’t feel productive. It’s to protect them from excesive waste and lack of investment in their technical debt
Also good to see working in transformation is even more valuable than tooling
This is one of the secrets of Pair Programming: it generally results in more coding time per person on average. (As well as that time being more productive.)
I love this data, the lens, and the way Adidas navigated this conversation.
They way they went about measuring everything really let’s them make evidence based decisions. Surprisingly rare to see because it’s sneaky hard to do. Awesome stuff.
An engineer's main job is to solve hard problems for the business not spending time on the keyboard.
That just made me so curious... "the teams that are not high performing is because we don't want them to be... they're good where they are"
Basically they work in areas of the company that require low change (ERP, some legacy integrations with scm…). So investment doesn’t worth the pain
100% the productive teams are the ones with more releases and open technologies
Improving ways of working and architecture is more important than GenAI 🙌
I think that more time-on-keyboard (TOK) for engineers is not necessarily a good goal. There is a lot of benefit to creativity in engineering as well as alignment activities. Reducing repetitive tasks is great, but I don’t like the idea of a simple metric to increase TOK.
I agree that more TOK isn’t 100% going to be a win. Still, there’s almost certainly some information content in the number. Outliers are probably interesting to investigate. Good metrics don’t give you answers, they allow you to ask good questions.
It reminds me of the notorious “lines of code” metric which led to bloated code bases
True… and yet when I found some developers had contributed zero lines of code for several weeks in a quarter there was value in knowing that.
Yep - I am certain that once some teams knew that someone was measuring “pure IDE” time, there was a lot more “activity” in the IDE!
Goodhart’s law at work. Which is why I liked @jchyip ‘s question about the correlation of IDE time with lead time (or other flow metrics).
I love it: “metrics are just there to create questions, they don’t give answers”
Exactly - “When a measure becomes a target, it ceases to be a good measure.”
☁️ Please welcome J*ohn Rauser, Director of Software Engineering, Cisco Cloud Security* and Anand Raghavan, Senior Director of Engineering, AI, Cisco. They're here to present Generative AI in the Enterprise: One Year Later
I'd love to see the Adidas store with a lens on flow state in particular. Time on keyboard has partial overlap but time at whiteboard is just as important. Has anyone looked at this in their own organization?
I like the comparison of using the AI assistants to working with/without Google.
A little wiring a little winning :first_place_medal:
These are the metrics that I love to see for Dev Productivity. 50% reduction on time for documentation and code auto-complete 30% reduction in repetitive tasks.
This is improving "potential productivity" or "perceived productivity" which absolutely makes a developer's live better. Where it breaks down is if the system around the engineer/dev team doesn't improve to allow for flow of value.
Harder to quantify: the amount of happiness that comes from those reductions in drudge work.
Still thinking about how to name the layers in a simple, clear way instead of the numbers.
This projection of AI on the three layers is 🔥. It resonates so well
"We delayed our product launch so we could take advantage of our internal AI platform" Short-term pain/delay, for longer-term enterprise gain. 🔥
“Learn from the customer.” … what a great place to start.
Literally, taking one for every other team. There must have been a product manager / P&L owner who might have been upset.
I hope the github copilot crew got some sleep last night. I have a feeling they are going to be popular today
“Make it easy to make it better” (make it easier for the customer to give you feedback to improve the product)
"Your average 10 year old firewall might have millions of rules." OMG.
"Shadowed rules. Duplicate rules." (This is a bit triggering.)
"That livestream is so addictive" --presenter Anand Raghavan regarding the live display of Slack messages on the main stage, which he was reading just before his turn to speak.
Yes! Our StageSync Slack-to-stage tool is amazing! Ping me if you’d like to have it at your conference. 🙂
Part of me is still gobsmacked that LLMs can apparently parse firewall rules.
"In product integration of thumbs up thumbs down" (making it easy to get user feedback)
I was just about to comment on how strong Anand’s shoe game was. . . take notes, @chris656
💡 IT Revolution should release an AI solution where we can chat with virtual versions of all the authors to get feedback on our ideas. That would be fun to explore. Virtual Jon Smart telling me to do better on my OKR would be pretty cool.
Haha yes, train the AI on the books, papers, and of course the videos, especially @jonathansmart1' CRAP (Certified Real Agile Practitioner) Lightning Talk.
Adrian Cockcroft created the repo with a concept to create a GPT for your own content https://github.com/adrianco/meGPT
Dora has something like that! @nathen.harvey mentioned it in his talks!
3We use it for solving semantic issues a lot, db migrations, connecting disparate vendors data structures and rapidly onboarding them etc.
:adobe: Say hello to Brian Scott, Principal Architect, Adobe here to present Generative AI Governance Strategy at Scale
Adobe: among many things, the creator of the PDF
Macromedia Flash was the peak of the Internet. The glory days are behind us.
real “get off the lawn!” vibes :rolling_on_the_floor_laughing:
Ebaumsworld conference is next week at the Tropicana iirc /s
I recall a radiolab podcast from 2017 covering an adobe gen ai product that could replicate a persons voice for text to speech with only 5-10 minutes of sample data. Very forward thinking
I'm hoping that someday, Brian will co-present with the SRE team that supports the insane Firefly product! What an adventure that must have been!!!!
“The system is down” by Strong Bad a classic https://www.youtube.com/watch?v=GtQpThwWQtQ
Props to @brian.l.scott recognizing Dan Neff as a partner in progress even though he wasn’t able to make it. 💖
"A funny thing happens: When you hang out with VPs and SVPs, and you're one of the few ICs (individual contributors), you always walk away with more work". 😂
how to account for the jerk (https://en.wikipedia.org/wiki/Jerk_(physics) )
I love this infographic for balancing these tradeoffs
Lesson Learned, don’t hang out with VPs or SVPs, Individual Contributors all the way!!
"Move as fast as a developer, with the certainty of a lawyer." Brian Scott Adobe .... #KillerLine
"We needed to move as fast as a developer, but with the certainty of a lawyer" 😂 (thank you!!)
💡 Thinking about the concepts of OSPO living near the AI legal, risk, and privacy concerns just opened a new door to explore.
Vendors slapping “AI” labels on their products and selling the same thing? :thinking_face:
hey they might be making an api call to OpenAI somewhere back there!
Two distinct use cases at Adobe: 1. internal use cases 2. external use cases
I missed it, was this an existing framework for evaluating risk and new processes at Adobe or something bespoke for AI use cases?
This was something new Brian created for these GenAI use cases.
1“We started to treat the process like a product.”
concern around sprawl continues to increase. Growing awareness of tradeoffs.
Love the use case model. Might be interesting to add some of the Value Canvas elements (from Jon Smart), e.g. outcome hypothesis & leading indicators.
They made a triage process for Security, Legal, and Compliance to streamline request process.
What does “shift left traffic light filters” mean in this context? Quality gates?
Awesome talk btw @bscott some fantastic highly quotable lines and heuristics
"it's been 15 years since I've given a talk... and might be another 15 years before I'm ever invited back" 😂
🔆 Please welcome Steve Yegge, Cody Platform Guy, Sourcegraph here to present: The Death of the Junior Developer, and Other Lessons Learned
“I am going to say something that gets me into trouble” - Says me every day
Free money, but less stuff (and then ppl wonder why prices are going up...)
"It was not until I went to Google that I knew what entitlement meant." Steve Yegge
Speaking through adversity. This may be my new favorite talk format!
Speaking through adversity. This may be my new favorite talk format!
The lingering gifts of the lock-down: entitlement and anxiety.
It is with great sorrow (and a bit of code cleanup) that we announce the passing of the beloved Junior Developer title.
"https://sourcegraph.com/blog/the-death-of-the-junior-developer" Steve Yegge
This is literally true: in my post following Yegge's post, >> PS: this is literally true: my wife/boss just told me she met someone who is a junior in computer science at Boston College, and asked me if I had any advice for her. I told her to email me, because I will send her a link to Steve’s post the instant it’s up.
Chat models suggest a LOT of BS code and you have to be deeply experienced to get “coding” value out of them. This seems very hard for people who aren’t fundamentally developers to understand
And I am pretty certain they suggest a lot of BS code is that they are “trained” on the current corpus of code - which contains a lot of very bad code.
Lead Engineer responsible for choosing what we DON'T include in the RAG
CHOP - CHat Oriented Programming - used by many senior programmers (but not junior ones -- yet)
Also literally true: I wrote: >> As a parent of a 16 year old, and two 14 year olds, how do I best prepare them for a world where entry-level jobs for new college graduates might be very, very different, and maybe even potentially far fewer in number?
if it takes a senior dev to do CHOP where will we get them if there is no longer a career path to get there?
In my opinion there’s a subtext here that’s missing: the time to become senior level is going to go down as well. (I mean what even is a senior engineer? That’s not an industry standard)
I’d imagine the qualities we look for in a jr. Dev will shift vs no new devs will be hired, right? Or are my glasses rose-colored?
Yeah I think there will be a subset of comp sci majors that are fantastic at reviewing code and others that are great at pure solutions (Faraday type thinkers) and it’ll sort itself out
The ones who squeak by just pounding out Java boiler plate will go into second careers etc
Right, and that sort of boiler plate code is something that frameworks like Spring were supposed to solve a while ago, right
Gosh yeah but the entire plugin ecosystem with spring just created an entirely different boiler plate problem
Pair Programming is now becoming reality with CHOP and help reap benefits of the concept introduced back in 1995.
Yes pair programming with a good assistant will be really powerful. I’ve also observed that people are much kinder to the LLM co-pilot than they are to human co-pilots!
Human co-pilots are so necessary! (speaking of airlines, with the death of pilots mid-air in the news recently)
Senior Devs are still going to be doing what Senior Devs are supposed to be doing: architecture, gluing different parts of the engineering business together, broadcasting best practices. Plus ça change…
"Coding is changing out from underneath you while you're still deciding whether to use it (Gen AI)"
Specialized agents are awesome. I have few of them that I use every day.
Sure. I have an agent to create user stories from a PRD. One for getting advices on architecture (AWS). One for helping mentor my mentees. One for meeting notes. One for implementing MVPs. One for writing Amazon style. And it goes on…
I’ll send this presentation to each of my engineers! 🧑💻 :female-technologist:
If AI generates all code starting now, will it basically train on itself in perpetuity until the singularity? Is The Matrix built on top of React?
Don’t Miss Gathr’s Demo at 12:20pm today at the Solutions Hub located in the Expo Hall. https://etlslasvegas2024.sched.com/event/1hAYU/solutions-hub-gathr-unleashing-genai-to-supercharge-your-development-lifecycle In today’s fast-paced tech environment, traditional management of DevOps, FinOps, SecOps, and other operational domains can lead to inefficiencies and competitive disadvantages. At Gathr, we believe GenAI is the key to unlocking a new era of XOps. Leading this transformation, our no-code platform empowers teams with flexible and intelligent solutions to enable automation and optimization. Join us for a dynamic 30-minute demo where we’ll showcase some use cases organizations have enabled to improve their teams’ performance, such as: • Conversational insights: allow teams to interact to and gain insights from their data using natural language • Intelligent pattern recognition: automatically identify trends and anomalies in performance results, suggesting targeted process improvements • Resource optimization: obtain inferences in metrics like infra usage and get recommendations for optimal utilization strategies, maximizing efficiency and minimizing costs • Proactive compliance and developer satisfaction: benchmark your work with best practices and review team sentiment to foster a healthier development environment Visit our Booth #305 or sign up for a free trial at xopsapps.gathr.one to see how Gathr can assist with your XOps initiatives and empower your teams to achieve more.
Don’t Miss Oso’s Demo at 12:50pm today at the Solutions Hub in the Expo Hall. https://etlslasvegas2024.sched.com/event/1hAYH/solutions-hub-oso-marie-kondo-your-authorization-the-life-changing-magic-of-tidying-up-permissions - “Life truly begins only after you have put your authorization in order.” – Sam Scott, Cofounder/CTO of Oso. Most people have grown accustomed to home-grown authorization systems that do not spark joy. Adding just one new role can take months; forget about adding custom roles. Then add performance problems and bug bounties full of IDOR vulnerabilities. The idea of a fully decoupled, centralized system feels like a distant dream. Join us for an interactive demo to see how to begin decluttering your permissions iteratively––without any migrations and with minimal refactoring. http://etlslasvegas2024.sched.com https://etlslasvegas2024.sched.com/event/1hAYH/solutions-hub-oso-marie-kondo-your-authorization-the-life-changing-magic-of-tidying-up-permissions View more about this event at Enterprise Technology Leadership Summit Las Vegas 2024
📣 And now, let's say hello to Patrick Debois, Dev(Sec)Ops Advisor & Author, here to present: Every AI Engineer Deserves an AI Platform, and Other Lessons Learned
"Our CEO said, we want every team to use this new technology"
"The people in the data dungeon... oh, they call it the data lake."
Is this like Lynyrd Skynrd getting sick of playing Freebird?
In the technology adoption lifecycle Patrick is an innovator. He can’t help but move onto new things!
Steps to scale the change: platform, enablement, governance 💡
In previous talks, @patrick.debois256 talked about how Marketing funded so much of R&D, because they were so eager to get something/anything to market.
Thumbs up and Thumbs down might not be an optimal feedback metric in most cases- people can do a thumbs down just because they had a bad coffee. A better way of assessing feedback is - if people copied the AI asistant’s response v/s if they click the Try Again button to get a different response.
Don’t pull AI adoption costs from your production budget if you don’t have a silver bullet use case
"After 6 months everybody hates frameworks" [...] "they're a thin veneer of abstraction". Because, this: "They're useful at first, but once you get a little bit under the hood, you're able to do it yourself." (better)
As someone currently building a framework :feelsbad:
I love that Patrick is emphasizing tests and the difficulties to overcome. This needs to happen to have stable, productionalized applications that support change.
YES! Code review times is going up, because more code is being generated. (I know of several pilots where they killed it, because it was putting too much burden on senior devs!!! Not tenable.)
One important thing to consider from what Patrick just said: AI bit rot is a real risk. Training RAG models on LLM generated code has poorly understood implications but so far it seems contraindicative
https://www.sciencedirect.com/science/article/abs/pii/0005109883900468
If learning fills the time we spent producing, how will that affect the “time in IDE” measure we were talking about…
Or put it this way… is the IDE even the right thing you want devs to “be in” in the future?
well if that becomes a measure, seems like it would preference an IDE-based “learning” tool :)
✨ Say hello to Paige Bailey, GenAI Developer Relations Lead, Google, here to present: How Google Is Radically Transforming Enterprise Software Development With Gemini
Thank you, @amandalewis, for making this connection!!!! 🙏
Great detail by Patrick! Excited to hear that him and John Willis are putting together a road show of hackathons and workshops to help people dig in.
As an engineer, @patrick.debois256 just summarized my anxiety with where we are in the hype curve of introducing AI into how we work and our digital products. I can't wait to see this space develop and move right on this journey. I'd recommend putting a chasm on this picture created by the lack of deterministic results. AI needs confidence more then deterministic solutions of our past.
Using Gemini Flash has been amazing — they recently lowered cost by 75%, and there's even a way you can use API for FREE. (They can train on that. But, the whole notion of "driving cost of intelligence to zero" is really happening.)
Can you share any metrics on the number of tokens alphabet is using internally? I’ve heard great things about the internal chat tooling
This is why CHOP is getting amazing — it’s driven by the foundation model advances.
It’s just a natural flow extension. It’s also a way of “documenting” or logging what would otherwise be your internal monologue.
I’m reminded of the TED Talk “We’re all cyborgs now”. This is going to be increasingly true as our machines have a memory of everything we’ve written and everything we’ve read and can not just answer questions but proactively prompt us to action based on that data.
In case anyone else was trying to learn more about CHOP: https://arxiv.org/abs/2404.01343
Amazing how big the "whole" of "coding" is — and the value when you've captured all of it. Telemetry, issues, service calls, code commits, compute, etc...
500k years of software development activity! But what % was “pure IDE” time?
Oh. Testing. I forgot about that!!! Cc @patrick.debois256 😂
26% of code written by code completions; 50% of code commits were machine generated?
Yeah, we find that chat programming generates a lot of code per query, so it bumps the numbers way up.
Are these code review features part of code Ai? Or planned for public release?
Google’s stuff is almost always internal-only, and I’m pretty sure their code infrastructure falls into that category, because it relies on some lower-level infra like high speed networks.
Dang, a constant topic of my 1 on 1s are how to generate more feedback on code reviews asynchronously.
I suppose this is where CHOP comes in. Is anyone posting their chat dialogues with PRs in your experience @steve.yegge ? That seems like it will become really relevant as we go further down the rabbit hole
In my experience, CHOP dialogs tend to be really long, thousands of lines. I haven’t had a chance to do conversations with the LLM during PRs yet, but I agree that it’s on the way!
Maybe just the links to the convo logs for audit logs? Please come find me as a new expert consultant in CHOP code reviews /s (your talk was the first time I had heard the term :) )
Gemini features are being integrated into Google Cloud, and already work via the APIs and AI Studio (if you ask Gemini to review your code :) )
On upgrades: one of my fave presentations was Jonathan Ju from Google, Java platform team, describing how difficult it was for devs to migrate from Java8 to Java 17 or something. Even Google devs aren't great at understanding heap space — the result was ballooning of memory usage, and tons of OOM errors in production.
Automated analysis of screen captures using product to identify friction points. For thousands of user sessions. 🤯
Gemini as Product Manager and UX designer! Nice - not distracted by “product vision” and instead driven entirely by reality.
The logical extension: ux transforms wildly as AI product designers begin designing experiences for AI agents
Love the “friction logs from user videos” to help capture UX improvement opportunities. Fascinating.
My mind responding with paraphrases of WOW!!! as Paige clicks through implementation of Gemini
Full disclosure: I usually do get Gemini to help me create slides (especially for design and images) :rolling_on_the_floor_laughing:
Google has the possibility to replace developers…based on a PRD! Not sure I am ready for it… god it’s a mind blowing presentation! 👏:skin-tone-2:
I guess it's finally time to be more serious about my carpenter side gig :robot_face:😅
I’m just hoping my Ai lottery tickets investments pay off so I can enjoy the glorious future our robot overlords will offer us
1@levi.geinert The domain is already registered, got beaten to it. I'm amazed and scared at the same time.
Personal plan: electrician, or coffeeshop + laundromat near my hometown in Texas :rolling_on_the_floor_laughing:
@webpaige Stop leading us to our doom 😅, it's not too late. I guess we'll need to start a support group for the reorientation of SWEs. But seriously, what you guys do is amazing!
Haha if Paige is worried now I’m worried!
🙌:skin-tone-4: Next up is Idan Gazit, Senior Director of Research at GitHub Next, GitHub, here to present Reaching for AI-Native Developer Tools
getting 85% there but struggling on the last 15% is a lament I've heard forever about frameworks & accelerators
I’ve also heard (and felt) the struggle with the first 15% - just getting started, where, how, … This stuff helps with that, too.
With punched cards, finding a program to identify LCM - it's in Aisle 5, Bay 21, Tray 3
I’m really enjoying the distinctions and differences between the different perspectives of these speakers
The notion of "this is so limiting" is so interesting — fascinating that http://Cursor.ai folks actually forked Visual Studio Code to escape the limitations of being "just an extension"
One could make the argument that VS Code needs to improve its extension APIs. 🙂
URL for cursor AI : https://www.trycursor.com/
God help all the IntelliJ plugin developers. 😂 I use the Cursive plugin, and the release notes complaining about breaking changes in IntelliJ are epic.
IntelliJ does have a pretty rich platform, actually a lot more open and capable than VS Code’s (though Java Swing, ick) — but they break backwards compatibility all the time and yeah it really sucks as JetBrains plugin developer.
That’s so super interesting!!! I have no experience with any of this, but suddenly, I’m imagining that IntellIJ gives you more access to the entire canvas, while VSCode traps you in windows or DOM elements? (My mind boggles that since last November, after years of almost no updates, it has new versions almost every week, and is broken half the time. I refuse to update it again, because it’s so aggravating when it breaks.)
"You get the worst outcomes when you ask for ambiguous things." <-- that's so helpful.
"prompt crafting"... sounds alot like "I'm so good at searching!" Our future devs are yesterday's internet deal shoppers.
"prompt crafting"... sounds alot like "I'm so good at searching!" Our future devs are yesterday's internet deal shoppers.
So much value is in the standard plan - review - build process that improves context.
The differing takes on “time in ide” remind me of Survivorship Bias: https://en.wikipedia.org/wiki/Survivorship_bias
GenAI needs to help us delete code, refactor, make it efficient.
I’d be curious if any of the tools we’ve learned about tackle this challenge!
Maybe we need to stop saying that AI will replace SEs. Maybe it’s just going to shift them from black ink (writing code) to red ink (reviewing code).
"the parlor trick of generating code" 🔥 We could steal so much from the TDD movement and think about codifying our spec as a prompt versus a test and work with AI to design, test, and implement. I want to see these tools play in the full TDD flow. Hot take, practicing true test first TDD makes me a better prompter
Some companies have started building a “semantic index” of their repos, where the LLM basically produces thousands of mini-PRDs describing the code base, which enables a bunch of stuff you could never do before. It’s absolutely 🤯 where they’re headed with it. I need to blog it up sometime soon.
That capability would be huge for legacy code bases that are difficult to understand.
They are fully internal to the two companies I’m aware of that are doing it, and their names are terrible right now. “Document store” is what one is calling it. Not very descriptive.
Hah. Everything is a Document Store.. I was thinking we might need to write a code analysis LLM that you can point at a legacy code base and it produce doc, mermaid, etc.
@internettitan but we just saw so much AI to improve/optimize code. Why do I need SE to review it?
1. AI generated code might never run with your existing code base. or never run at all... 2. Improvised code might improve one aspect while severely affecting/degrading other aspects of your software/program
In the 1990s everyone said the same thing was going to happen to animators when CGI arrived… Didn’t happen. In the 2010s everyone said the same thing about ops professionals when devops/automation became mainstream…. Didn’t happen. What makes this time so different? How will this unique time/context break Jevons’ paradox - increases in the efficiency of a resource actually lead to an increased rate of consumption of said resource through increased demand (e.g., when higher fuel efficiency and lower emission cars are sold, total driving and emissions go up, not down).
This is a really great idea, using the tools to move toward literate programming, which we’ve never been able to pull off before.
Reminds me of TogetherJ #iamold
Is auto documentation problematic if it is a lie? :face_with_monocle:
“We’re helping people not just with their typing. We’re helping them with their thinking.”
If you pair program, you often articulate all your thoughts out loud. These are navigation prompts for you and your pair partner. Make it so I can do this out loud with my voice. Join the discord between me and my pair, don't interrupt it with chat.
What did punch-card programmers think of IDE + Search Engines as they arrived? Software Engineering is here to stay, the tools are growing so 1 FTE can cover more ground.
💡Next up is Rosalind Radcliffe, IBM Fellow, CTO for CIO Technology Platform Team, IBM, here to present: Hosting Our All-Company AI Challenge
178,000 participants in a hack-a-thon… that’s a lot of pizza.
I propose you order as code...https://registry.terraform.io/providers/MNThomson/dominos/latest/docs
Jokes aside...what we learned from search engines is that it is a short path to the BOT teaching you how to talk it's language to get results, while you pretend you are good at teaching it what you want.
"Only" 178,000 people participating. 11,500 projects submitted, all judged! Love how IBM was trying to invest in all it's humans learning, upskilling with this effort.
walking back from break with Rosalind, I mentioned (with some pride) that we were judging 26 submissions for the EA awards currently. She very graciously did not mention these numbers to me.
"Challenge extended to Saturday because we broke the system" (capacity issues!)
Love the customer [0] model. Getting your own company invested in testing/breaking/using your own products is always worth the time and effort.
Imagine if hackers used their systems thinking skills for organizational problems... So much potential.
"As a matter of cosmic history, it has always been easier to destroy than to create." -Spock
0% chance the attackers stopped at one project. The initial infiltration happened several years ago.
Exactly... how many projects do they already have in their pocket, contributing value until they can quietly let something slip by...
An unsettling question: what if OpenSSH wasn't actually the real target.
I think Debian build servers were completely rebuilt, because it was assumed compromised.
One of the most fascinating parts of XZ was the exceptionally high quality of the commit messages introducing the exploit
For XZ it took a while to develop the social trust to push through the expoit... the social hackery is quite clever
I love how Dr. Ethan Mollick put in invisible text on his website, which now shows up in ChatGPT.
Sometimes the act of explaining the theory of an attack can become a roadmap to implementing a version of the attack…
The persistent attacker will learn the theory anyway. Raising awareness helps uplift the rest of us.
More likely a nation-state, typically not really interested in self promotion.
Not the institution.. but the individuals actors... well, at least they get a mention in the classified internal history of the org...
@nickeggleston just search LinkedIn for the resume that takes credit. Maybe ZeroCool did it
OMG. Same thing was attempted on jquery, loads, electron (?!?)... wasn't successful, but whoa. Like all single point of failure for all the JavaScript ecosystem.
Strength comes from the community. > Friends don’t let friends commit backdoor code changes.
Single maintainer + pressure ...ouch. No peer review, no oversight. OSS fears 101 #WhyWeCantHaveNiceThings
Single maintainer + pressure ...ouch. No peer review, no oversight. OSS fears 101 #WhyWeCantHaveNiceThings
The inherent transparency of OSS gives opportunity to find this stuff too. But yeah, it's pretty terrifying.
Teams of AI agents constantly trying to hack our own systems. A security Chaos Monkey if you will.
Increasing transparency of code is good, but also better authentication/validation of contributors is helpful. How can you eliminate anonymous github user accounts?
Reminder: Please submit your feedback for the talks you attended. It’s so valuable for us and the speakers. And after all, feedback is a gift and sharing is caring! Enter your feedback for those talks here: https://etlslasvegas2024.sched.com/ https://devopsenterprise.slack.com/files/UATE4LJ94/F07HGRF5J2G/feedback.png
Could the OSS value proposition be dead vs the risk if I can rapidly produce reliable, performant, optimized code with AI when I need it?
Don’t Miss Gathr’s Demo at 12:20pm today at the Solutions Hub located in the Expo Hall. https://etlslasvegas2024.sched.com/event/1hAYU/solutions-hub-gathr-unleashing-genai-to-supercharge-your-development-lifecycle In today’s fast-paced tech environment, traditional management of DevOps, FinOps, SecOps, and other operational domains can lead to inefficiencies and competitive disadvantages. At Gathr, we believe GenAI is the key to unlocking a new era of XOps. Leading this transformation, our no-code platform empowers teams with flexible and intelligent solutions to enable automation and optimization. Join us for a dynamic 30-minute demo where we’ll showcase some use cases organizations have enabled to improve their teams’ performance, such as: • Conversational insights: allow teams to interact to and gain insights from their data using natural language • Intelligent pattern recognition: automatically identify trends and anomalies in performance results, suggesting targeted process improvements • Resource optimization: obtain inferences in metrics like infra usage and get recommendations for optimal utilization strategies, maximizing efficiency and minimizing costs • Proactive compliance and developer satisfaction: benchmark your work with best practices and review team sentiment to foster a healthier development environment Visit our Booth #305 or sign up for a free trial at xopsapps.gathr.one to see how Gathr can assist with your XOps initiatives and empower your teams to achieve more. (edited) http://etlslasvegas2024.sched.com https://etlslasvegas2024.sched.com/event/1hAYU/solutions-hub-gathr-unleashing-genai-to-supercharge-your-development-lifecycle View more about this event at Enterprise Technology Leadership Summit Las Vegas 2024
Hey all! Want to talk about GitHub things or AI things or GitHub AI things or really anything? :thinking_face: Come on down to Capri 2 and let's talk! 🐙😸
excited to present after lunch ! please come say hi afterward, i'll be standing around outside
Don’t Miss Oso’s Demo at 12:50pm today at the Solutions Hub in the Expo Hall. https://etlslasvegas2024.sched.com/event/1hAYH/solutions-hub-oso-marie-kondo-your-authorization-the-life-changing-magic-of-tidying-up-permissions - “Life truly begins only after you have put your authorization in order.” – Sam Scott, Cofounder/CTO of Oso. Most people have grown accustomed to home-grown authorization systems that do not spark joy. Adding just one new role can take months; forget about adding custom roles. Then add performance problems and bug bounties full of IDOR vulnerabilities. The idea of a fully decoupled, centralized system feels like a distant dream. Join us for an interactive demo to see how to begin decluttering your permissions iteratively––without any migrations and with minimal refactoring. http://etlslasvegas2024.sched.com https://etlslasvegas2024.sched.com/event/1hAYH/solutions-hub-oso-marie-kondo-your-authorization-the-life-changing-magic-of-tidying-up-permissions View more about this event at Enterprise Technology Leadership Summit Las Vegas 2024
@genek are all these talks recorded? I want to rewatch some of them. Will they go up on YT or somewhere?
Absolutely — they’ll be posted at https://videos.itrevolution.com/ “soon” — days, not weeks. Plenary talks likely first and faster. One more fantastic reason to get an ITREV membership! 🎉🎉🎉
And now, get ready for George Proorocu, IT OPS Chapter Lead - Cybersecurity & Fraud, ING Bank, here to present Deep Fakes: The Lies We Can’t See 😱
A great way to start the afternoon “How AI is used by bad people to do evil things.”
Trend moving from "scam grandma or parents" to "scam enterprises"
Link to the Ferrari CEO impersonation: https://x.com/vxunderground/status/1818339469268832541?s=12&t=bXOSUjHCmTqhRMcuRxs-tA
Anyone notice that @aurel-george.proorocu seems, umm, really good at delivering these social engineering script? 😂
Seems like just yesterday that "is your refrigerator running" was the worst phone call we could get. Wow.
Wondering when we’re going to get regular testing for these type of scenarios like we do currently with phishing tests.
I keep thinking about something @aurel-george.proorocu said to me at the speaker recpetion: “we will all be probably making more changes to approval processes than we have in the last fifty years.”
Who said it is hard to find a proper business case for AI?
I’m sitting next to Tom Cruise… he’s shorter than I thought.
Must practice these high pressure authentication scenarios
How accurately can AI determine if the video/audio is a deepfake? AI vs AI
🌟 Please welcome Shawn "Swyx" Wang, Writer, Founder, Devtools Startup Advisor, here to present Rise of the AI Engineer
@aurel-george.proorocu curious what software you used to create those videos. No reason. Just curious 😉
Facefusion on Git for the video demo - https://github.com/facefusion/facefusion. http://Elevenlabs.io for the audio deepfakes.
1“I’m trying to start my own industry.” This is a brilliant 4D-chess-level strategy.
None of them know how many rs are in strawberry
Cost of entry in the LLM game: $100MM data center build (Microsoft); $25MM for data sets (Meta)
$/intelligence race: so real. Gemini Flash API access is free. I had to read that a couple of times — couldn't quite believe it.
“Ready --> Fire--> Aim: ‘Agile comes to AI’” great slide title.
All sorts of economics change when $ per token goes to near-zero.
of course, how much of the cost per token is being artificially lowered by VC funding?
Thanks to Shawn’s article I now have 3 AI Engineering Xperts on my innovation team at Excella
AVOps in the back deploying a fix fast “Mouse cursor to the X”. Accelerated deployment
Some of the greatest art humans have created came from "hallucination." We should not stifle this new source of creativity.
👏:skin-tone-4: Welcome John Willis, here to present Dear CIO: Navigating the Shadows – GenAI’s Promise, Peril, and the Path Forward
I guess it has seemed historically that “AI” was an umbrella concept that included machine learning, computer vision, and, now LLMs.
John Willis was in the room during that famous Allspaw/Hammond "10 deploys/day at Flickr" presentation at in 2009 — he mentioned, "today felt like THAT!" (What an interesting observation.)
Grab that paper here: https://itrevolution.com/product/autonomous-ai-in-the-enterprise/
3
For @stephen and me, the "P" in LAMP stack is Prolog.
“Ai is just network, compute, and storage. The icing on that cake is looks good, but the inside is the same.”
It was wild to see people using ChatGPT Code Interpreter to get access to the underlying OS, probe the kubernetes cluster, etc. Sandbox definitely was a little leaky, in the early days. In fact, I think @shawnthe1 live streamed doing this right after some OpenAI launch.
Make AI agents as easy as a lambda. I agree John, this is an infrastructure problem.
im outside the Azure Ballroom if anyone wants to chat! slides here https://docs.google.com/presentation/d/1SqLdUmhyRz1aEOWEnlqsDCGp0r-32iKhT8eSS0XWBwM/edit?usp=sharing
⭐Please say hello to André Martin, Founder, Shift Space, LLC, here to present Ready, Set, Lead: Maximizing the Space Between No Longer and Not Yet
this is what my kid says when they want to skip school. 😆
"You know everything — be our teachers." "I still don't know how to quit emacs."
There are so many "Masters of Craft" here at this conference. I'm so glad to be back this year.
https://www.menti.com/aloarqzc814e?source=qr-instructions-widget
I want to keep the connection to this "scenius" after the conference ends, to keep growing by the experiencing of challenging each other.
https://www.menti.com/aloarqzc814e?source=qr-instructions-widget
Inspiring call to growth by Dr Andre Martin... now, how to operationalize it...
Gen AI has the potential to do profound damage if it's not done right.
News! All messages with 💬 reactions will now be automatically posted to to the #C07HM6V4P2A channel (thanks to @jeff.gallimore)
Your calendar: Where you spend your time dictates what’s important for your company.
What is your decision tree for what you say "yes" or "no" to when letting time get booked on your calendar?
I've started living by this mantra "You are the CEO of your time. It is your currency; spend it wisely" - Tim Muttitt -- Michael Scott - Tim Muttitt
Let’s keep developing ourselves in these arenas (while we teach the world about GenAI).
@brunneradamj The idea behind “craft circles” is to create a space to practice, play around, and get comfortable while we build stronger relationships and gain from each other’s journey into GenAI. Imagine having a session about how to use GenAI to do XX, invite people to the space and allow them to practice, play, share, freefall, fail, succeed, etc.
I think we should have a breakout for fellow travelers that deal with high sev Prod incidents during ETLS 🥴
now you get the sigh of relief that it's working and begin to deal with the fallout!
No frosty’s at Wendy’s and high out of stocks at Target… sigh
@david.faircloth and I are going to need a cocktail on GitHub and Liatro tonight
Reminder: The breakout sessions are starting in 5 minutes. Start navigating your way to whichever session you’re attending. https://devopsenterprise.slack.com/files/UATE4LJ94/F07H7N10D62/timer.png