Peter Bell

Author @OReillyMedia "Scaling AI Adoption in Eng" Founder @Gatherdev the CTO community for CTOs who don’t have time for community. Host CTO summit @Kubecon_

[Peter Bell @PeterBell]

Agreed. I'd say the (competent) innovators will continue to get alpha from composable harnesses and being on the bleeding edge, slowly replacing elements with OSS or commercial alternatives (e.g. temporal for durable execution).

Early adopters and early majority should be using novel commercial tools to accelerate learnings (like yours!).

Late majority and laggards will just wait until it's added to teams 365 and copilot

[Peter Bell @PeterBell]

I'm all down for one shotting when it works, but there is a real place for rigorously describing requirements and/or verifications. These days you can start semi structured, but eventually I think you iterate into a structured set of DSLs for the important verifiable constraints

[Peter Bell @PeterBell]

For anyone who missed it on @hackernews Dijkstra reminds us prompt engineers that those who do not learn from history are bound to repeat it - on the foolishness of natural language programming https://t.co/BGimHaJicL

[Peter Bell @PeterBell]

NO question - almost all of the value from AI comes when you rebuild your SDLC.

Heck - chatted with a very large comnpany last year that reduced cycle time by 60% smply by getting agents to capture rigorous requirements from the business team upfront and not stop until they were sufficiently complete. Saved all of the "hold the line while we book a call with an SVP to understand what they meant by 'customer'" stuff. It's not just about generating code faster

[Alex Kemper @ajkemps]

More here: https://t.co/yUYOsxUQIe

[Peter Bell @PeterBell]

Hey @ajkemps I don't disagree and have seen that too, but it's bifurcated. Shopify has built their own thing, as have block, ramp, coinbase and many more, I know a number of other CTOs that have composed great factories tuned to their needs. It's not easy, but it's not hard.

So saying some companies have done a bad job is right. But that's on the team, not the idea of composing vs buying.

[Peter Bell @PeterBell]

He's not raising, but you should ask @GeoffreyHuntley for his opinions on Unikernels https://t.co/MNemrzPmck - one small part in reducing attack surface area for AppSec - and lean and other formal proofs will become more important.

Beyond that the ability to share identity and a subset of approved permissions with your agents - it's ridiculous that websites want to check if I'm a bot - I don't want to go to a website, I want my duly authorized agent to go to it. Looking forward to seeing that get widespread adoption. Interesting times!

[Peter Bell @PeterBell]

I compose my own systems because (a) I want to own the context and tune the dreaming and harness and (b) because none of the off the shelf solutions are terribly good yet.

What is your story around multi-player mode (20 developers driving specs and verifications for hundreds of agents), your opinions and approaches to rigorous (deterministic) verification of correctness, performance, security and maintainability, what's your loop for reviewing local telemetry to tune resolvers and skills based on failures, how do you pull evals from prod workloads to select and experiment with alternative models and hill climb content/skill tuning for each to allow models from different labs to be used for the same steps, what's your fail over capacity if a given lab is down, and what are you doing to shift testing left so agents don't have to wait for CI which is way too slow? What does your shared capacity broker look like for managing limited resources like CI and potentially staging environments, what are your flows from operational telemetry for both auto bug fixing and providing context for an automated PDLC based upon user behavior . . . I'm running solo and these are my baseline expectations.

[Peter Bell @PeterBell]

Amazing night last night at DirectorPlus in NYC - thanks so much to the speakers, sponsors and participants - I'll tag andf thank everyone on LinkedIn, but a great night discussing agentic engineering and security. https://t.co/zJdmMQgkI6

[Peter Bell @PeterBell]

Welcome to a taste of what a random Wednesday is going to look like every week in 2027

Time to look at AppSec and cycle time to patch again. As long as you can go from CVE in inbox to tested patch in prod in under 10 minutes any time of the day or night, you're probably fine...

[⁵⁄₉ @fiveoutofnine]

So basically, today PHP got two critical bugs that let anyone run code on your server without even logging in, which is bad, and then the details leaked before the patch was even out, which is worse, and it works on WordPress and MediaWiki and Magento and Joomla and Laravel, and Censys says like 1.5 million servers are running a version with no fix, so the NSA was like "take your system offline," and the FBI was like "take your system offline," and CISA and the UK and CERT-EU were all like "take your system offline," and then the Prime Minister of Japan, the actual Prime Minister, posted at 5:55 in the morning telling 48.8 million people to take their system offline, so everyone took their system offline, like the Dutch cybersecurity agency took its own website offline, and the Dutch parliament went offline, and the Port of Rotterdam went offline, and Finland and France and Denmark started pulling government sites, and Wikipedia went down, and https://t.co/QlOXb3COq3 went down, and Asana and GOG and Something Awful went down, and the University of Vienna's webmail went down for some reason, and that's just PHP. Because GitLab also dropped a CVSS 10.0, which is the highest score they give out, for a bug that's been there since 2014, so that's every self-hosted GitLab for twelve years, and then LiteLLM 1.107.0 got hijacked and shipped a credential stealer and a remote access trojan in every single wheel and the Docker images, including latest, and you didn't even have to import it, just installing it was enough, and it's the second time this year, and then Railway paused every deployment on the entire platform because someone got into the environment variables, so now everyone has to rotate every secret they've ever stored there, and TiDB Cloud said someone was walking around inside with stolen credentials for basically a month and an extortion group says they took 10.6 terabytes, and Zeabur got breached for the second time in two months and is just shutting down, like the whole company, and Node.js patched remote code execution and a TLS bypass, and CISA says Plex is being actively exploited, and someone took over 300 Tor exit relays to steal SSH passwords, and SafePay ransomware has hit over 600 critical infrastructure orgs, and India's biggest bank filed a data loss incident, and Japan has been in an official state of emergency in cyberspace since October 9, which honestly explains the 5:55 AM post. Oh, and eMudhra, a certificate authority that's issued over 60 million certificates, whose entire job is proving websites are who they say they are, was supposed to be using private keys that are random numbers over a hundred digits long, and one of them was 4. Just 4. And the others were like 66 and 237. Anyway, it's Wednesday.

[Peter Bell @PeterBell]

@patio11 @allTheYud Ohhhhh LLMs - they haz opinions. Especially when another LLM threw a half baked idea out a week ago and recorded it as a decision because you didn't say no . . .

[Peter Bell @PeterBell]

@mikehostetler Not sure I care. For me I actually am fine with note takers - I just expect professional courtesy and respect for frieNDAs. I don't care whether you remember stuff in your brain or in a database, only that you don't misuse it and don’t allow training on it

[Peter Bell @PeterBell]

If you're in SF on 11/4 and want to catch @GergelyOrosz along with @CompleteSkeptic and @bcantrill you should sign up for this! Thanks to the @WorkOS team for hosting!

https://t.co/kMiqqx4wHA

[Peter Bell @PeterBell]

@GeoffreyHuntley I have something I use for Anthropic @GeoffreyHuntley but I'm running each anthropic plan on a separate user account on a machine with a UX (not headless). Injecting credentials in a single user account broke in the late spring and I lost a morning rebuilding it all.

[Peter Bell @PeterBell]

@duncan @GeoffreyHuntley Its enough to make someone take up Photography to calm down :)

[Peter Bell @PeterBell]

Looking forward to giving this a go - thanks @mikehostetler

[Mike Hostetler // Actors & Agents on the BEAM @mikehostetler]

Wait, you cut lines of code by 20% and doubled the features?

Yup, that's Jido V3

Previews begin this week

[Peter Bell @PeterBell]

@boardyai @hf0 30

[Peter Bell @PeterBell]

My theory? The Roadster by @elonmusk is an homage to @jack and the twitter of old . . .

Twitter: "you can post anything - as long as it's only 140 characters"
Roadster "you can fly anywhere - as long as it's under 300 yards"

:)

[Elon Musk @elonmusk]

Turns out the future does have flying cars

[Peter Bell @PeterBell]

Hey @GeoffreyHuntley I hope you have a bot blocking the reply bots? :)

Not top of mind for me, but I'd imagine it'd be a fun excuse to play with jev as a classifier and depending how grump you feel you just change your confidence level for blocking.

Maybe even post it on your x every morning or your human API

I just want to see the day where you go all Dirty Harry and block at confidence of 0.01 "You've got to ask yourself one question: 'Do I feel lucky?' Well, do ya, punk?" :)

[Peter Bell @PeterBell]

@GeoffreyHuntley 👀