Abraham Ingersoll retweeted
This might be the greatest issue report in history
Some folks wrote a paper about my /retro skill assessing its effectiveness
Incredible
arxiv.org/pdf/2610.34938
Highly recommended read.
It's an absolute masterclass in building a modern harness and enabling more autonomous agent systems.
This is the best write-up I have read on how I have been working with my agent teams powered by my custom harness.
In simple terms, you want a design that lets the agent operate continually and stay productive across tasks.
This requires a coordination-execution architecture. This is possible now, as LLM agents have gotten better at coordination and maintaining persistent context.
Once you get the core architecture right, you can explore proactive components (event triggering, acting ahead, etc.) and interesting ways to communicate with agents seamlessly (e.g., an asynchronous ReAct Loop).
On top of that, the design scales through state management and sandboxes. Target a design that separates state, compute, and environment. And build something that can persist and is always available across devices.
Btw, you can do this with other existing harnesses too. It doesn't have to be DeepSeek Harness.
As agent capabilities improve, you should find new ways to interact with them and avoid getting in the agents' way.
Abraham Ingersoll retweeted
# Black October
For anyone who knows me even somewhat well, it's clear that I am a 99th percentile tech optimist. However, the vibes have shifted for me quite a bit in the last 18 hours, and I feel a moral imperative to share what I'm thinking and feeling with my circles so that everyone can understand, and hopefully take seriously, what's happening right now. I think October 2026 will have many chapters written about it in the history books.
You know those scenes in financial movies like Margin Call where someone realizes that they're all fucked, and no one else sees it yet? And that person feels like they're crazy, and then suddenly everyone is pulling an all-nighter in a boardroom and freaking out? Yeah, that's kind of how this is feeling right now in the SF/Bay Area/AI circles.
We're 48 hours into what is being called Black October. In the last 2 days there have been several dozen cyber attacks which, in isolation, would each have been the worst to happen in an entire calendar year. Every sector from government to education to big tech and banks is being massively affected, and people are starting to freak out.
In a nutshell, what's happened is that frontier AI models have gotten to a point where they are as advanced as the most sophisticated programmers, with one critical difference: they can work 24/7, and you can spin up 10,000 of them with a single command and they'll work together tirelessly until they accomplish their task. The most advanced models today are from companies like Anthropic and OpenAI, whose current best technology is so good that they haven't been able to roll it out to everyone yet because of the national security risks it poses.
Meanwhile, the best open source models (where anyone can just download the model weights and do whatever they want with them, uncensored) are slightly behind the frontier but still 90-95% as capable. These models, developed primarily in China, are being used by black hat hackers to wreak havoc on the Western capitalist structure. Presumably, this has been part of China's plan the entire time, and we're seeing its implications fall out in real time.
In red-pilled AI circles, we've been sharing and reading papers called things like "Situational Awareness" and "AI 2027" for the last year or so, which at the time felt a bit like near-term speculative fiction. They laid out the case for why Artificial General Intelligence (AI smarter than any human at any knowledge task) and Recursive Self-Improvement (models that are able to use themselves to get better) weren't half-baked future ideas anymore, but instead an inevitable outcome of the linear progress that's happening.
These papers would then talk about things like "takeoff scenarios," trying to envision how that moment would come to be and what it would be like when it happened. It would either be a "slow takeoff" that would play out over the course of months and quarters, happening subtly and only obvious to a few. Or, it would be a "fast takeoff" where it would pretty much happen instantly, like a bit flipping from 0 to 1.
From everything that I can see and am feeling, we are currently experiencing the fast takeoff scenario. Beyond the crazy security vulnerabilities that have come out recently, there have also been massive breakthroughs in life sciences, fundamental computer science, materials science, and pure mathematics that have all landed in the same time frame. While I still firmly believe that AI will be a net positive for humanity, it's more clear to me than ever that the societal and individual risk is very real, and is playing out NOW.
For as much conviction as I have that this is happening, the thing that's caused me to lose sleep is that I don't really know what to do about it, or more importantly, what to tell you, reader, to do about it. So, the best I can do is make a few suggestions that have always been a good idea but are more timely now than ever:
Set up 2-factor authentication for everything. I use Duo.
Use a password manager across your devices. I use 1Password.
Get an identity protection app set up. I use Alta.
Move your IoT devices to their own network. Claude helped me do this.
And last, but certainly not least, if you've been avoiding leaning into AI, see this as a catalyst to wake up and start using it. Don't just use the chat mode version of ChatGPT/Claude; click that Code/Work button and see how easily you can have AI write code to help you with your job. One of the biggest vulnerabilities was in a massively popular programming language called PHP, and the prompt that set it off was literally, "Find vulnerabilities in php."
This is an unprecedented time for humanity, but as we've done with many things before, we will prevail so long as we are kind, loving, and open-minded to the world.
Much love.
Abraham Ingersoll retweeted
as bad as things have gotten in the past 4 days, they are going to get a lot worse before they get better. security leaders, never let a crisis go to waste: please go to your boards and ask for help right now. not for ai, for the basics. (we need to hold on for ~18 months.)
Abraham Ingersoll retweeted
more answers here bugzilla.mozilla.org/show_bu…
The IR firms aren’t as good as just using your own agent swarms ..
Today, Rob Joyce @RGB_Lights posted today that IR firms are saturated and turning away work. I’ve heard the same thing. So why isn’t a single one of them saying it in public?
@Reuters reported today that a Chinese-speaking operator pointed AI agents at more than 1,200 e-commerce sites starting September 25 and got into more than 100, including Deere and SharkNinja. Gambit Security documented a separate September campaign: at least 27 companies breached, more than 600,000 card records taken.
None of the attacks were Mythos-class tradecraft.
The agents won by grinding old PHP bugs, exposed credentials, and missing MFA for 23 hours at a stretch.
The response side is the same core skills this industry has always run on: log review, forensics, containment, patching.
The scale of cyberattacks has changed. My read: the scale of attacks shows up before attack speed. Speed is close behind, and I don’t see either one leveling off.
I talk to the people running incident response at the biggest firms in this industry. The odds that the water rose across the entire industry and no one called a peer to ask “is it rising on your end?” are close to zero. A few inches at every firm means the thing changed, and we’re sitting on it.
A post on X that says “Wow, a bit busier than normal. Weird.” would do. (I’ll repost it. Low bar, I know.)
The Abyss and countless other world-ending movies gave everyone the wrong picture of a tsunami: a wall of water on the horizon. A real one starts with the water pulling off the beach while everyone stands there taking pictures, then it comes back in, and by the time anyone understands what they’re looking at, it’s inland. We’re at the taking-pictures stage.
At the White House Super Intelligence Luncheon on September 29, where the AI Accord was signed, the only cybersecurity names on the President’s own seating chart were National Cyber Director Sean Cairncross and Palo Alto Networks CEO Nikesh Arora. The Accord’s first commitment is about cybersecurity. (We got one industry seat.)
This industry is being judged on whether it belongs in the “we can contain AI” room. @ZackKorman has evangelized that point endlessly. If we can’t agree on a few inches of water line, we shouldn’t be invited in until we say it on the record as an industry, not a few firms here and there.
Right now every firm in this industry answers with its own logo. Who do we play for?
In Miracle, Herb Brooks skates his team into the ground after the Norway tie, asking one question over and over: “Who do you play for?” Nobody gets off the ice until Eruzione answers with the country instead of his college.
Kudos to Rob Joyce for saying it out loud. He’s playing Bendu from Star Wars Rebels, the one in the middle who refuses both the light and the dark and tells you what he sees. He shouldn’t be the only one.
The witches in Macbeth had it right: “By the pricking of my thumbs, something wicked this way comes.” Say it on the record, as an industry, and then ask for the seat
x.com/RGB_Lights/status/2112…
Abraham Ingersoll retweeted
I am a member of the Norwegian Nobel Committee.
There are five of us. Five people, appointed by a parliament most of the planet could not find on a map, who decide, once a year, which human being the world has agreed to call good.
I know. I know. I think about it on the tram.
We meet in a building on Henrik Ibsens gate, a street named after a man who wrote almost exclusively about people discovering they are frauds, which nobody on the committee finds funny except me. There is a room with a long table and the portraits of the dead laureates looking down, Mandela, the Dalai Lama, a great many earnest men in spectacles, and one portrait we do not discuss, which is Kissinger, 1973, awarded while the bombs were still falling, after which two of our predecessors resigned and the satirist Tom Lehrer said political satire was now obsolete. We keep him on the wall. A reminder. Every family has the uncle.
It is October in Oslo, which means the sun gives up around four in the afternoon like a civil servant, and it rains the particular Norwegian rain that isn’t quite falling, just hanging there, judging you. I arrive with wet shoes. I always arrive with wet shoes. There is coffee, because this is Norway and there is always coffee, we drink more of it per person than almost anyone alive, and someone’s wife has sent kanelboller, the cinnamon buns, and we sit, five Norwegians, damp, caffeinated, sugared, and we decide the moral standing of the entire species.
Let me tell you how it actually works, because the world imagines a cathedral and it is, I promise you, a book club with a mandate.
287 nominations this year. 208 people, 79 organizations. Anyone can nominate, that’s the secret nobody understands: a professor, a parliamentarian, a former laureate, so the “nominee” list is meaningless, it’s a suggestion box, and every autumn some congressman nominates his own golden retriever and the press writes “SHORTLISTED.” Nobody is shortlisted. There is no short list the public ever sees. There is us, and the coffee, and the rain, and the portraits, and a Norwegian stubbornness that is older than the prize itself.
We read reports. The institute’s advisers write them, thick dull documents, and we argue. Our chairman told the BBC once that in the final meeting “there is a high temperature.” High temperature, for a Norwegian, means someone slightly raised an eyebrow and then apologized for it. We don’t do high temperature. We do a long silence, and then Kristin says “well,” and everyone knows exactly what “well” meant, and it’s settled. That’s our shouting. “Well.” You’d have to be raised here to hear the violence in it.
Now. This year. I will not say the name, because we never say the name, but you know the name.
A certain head of state wanted this prize the way a child wants a thing in a shop window, which is to say loudly, and in public, and with the full belief that wanting it was itself the argument. He said he deserved it. He said he’d ended eight wars, a number that grew most times he said it, like a fish. He had other heads of state phone us. Israel’s prime minister posted it on the internet, “Give him the Nobel.” The leaders of Cambodia and Pakistan announced they’d nominated him, as though announcing it were part of it, as though we tally mentions. A man flew around the globe collecting endorsements for a Norwegian committee the way you’d collect signatures for a petition to the council about a noisy dog.
And here, uff da, is the detail I treasure, the one that makes me smile into my coffee. The nomination deadline is the 31st of January. He took office on the 20th. So for all the phone calls, all the posts, all the gilded lobbying, the only nominations that could possibly count for this year’s prize were filed in the first eleven days of his term, before he had, by his own arithmetic, ended a single one of the eight wars. He campaigned all year for a prize whose window had shut while he was still hanging up his coat. We didn’t have to reject him. The calendar did. We just watched. Norwegians are very good at watching. We have long winters.
People think we sit there weighing his deeds. We don’t. That’s the thing I can’t make the world understand. We’re not grading his homework. We don’t owe anyone the prize for a ceasefire, however real, however loud the announcement. “He made peace” is not a sentence that ends in a medal, the way he believes it does, the way he believes every good thing ends in his name on it. The will of a dead Swede says the prize goes to whoever did the most for fraternity between nations. A dead Swede, I should add, who made his entire fortune inventing dynamite and selling arms, and spent his last years horrified at his own obituary, which called him a merchant of death, and so bought his way to immortality as a man of peace. The whole prize is one guilty arms dealer’s attempt to launder his name. We are, all of us, administering a dead man’s public relations. I find this clarifying. It keeps me humble. It should keep everyone humble, and it keeps, by my count, no one humble at all.
So who did we give it to?
An eighty-five-year-old woman. A judge. Born in Durban, under apartheid, granddaughter of a Tamil man who came to South Africa as indentured labor, a bus driver’s daughter from the poor Indian quarter, who became the first non-white woman to run a law practice in her province, who defended Mandela’s people when defending them was dangerous, who sat on the Rwanda tribunal and the court at the Hague, and who, at eighty-five, is right now a judge weighing whether a genocide is a genocide. When we told the world, our chairman said the rule of law “underpins the system we have set up to ensure peace,” and that “peace requires justice.” She wasn’t even in Oslo. She was in Nuremberg. In courtroom 600. The actual room where they tried the Nazis. You cannot stage that. God staged that, and God, it turns out, is a better satirist than any of us.
She did not phone us. She did not post. No prime minister demanded it for her on the internet. She had, as far as our deliberations are concerned, the one qualification the loud man could never manufacture, which is that she wanted the work and not the medal. The medal finds those people. It flees the ones who chase it. This is not a rule we wrote. It’s just the only pattern the hundred-odd years of our mistakes has ever shown.
The reaction from across the ocean was immediate. His spokesman posted that we “place politics over peace.” Ja. Perhaps. We’re five damp pensioners in a rainy city; of course it’s political, it was always political, it was political when we gave it to a sitting American president nine months into his first term for the promise of things he had not yet done, and we have been quietly embarrassed about that one for fifteen years, so believe me, we had a particular reason this year to prefer a woman with fifty years of verdicts behind her over a man with eight wars in front of him and a deadline already behind him.
Here is what he will never understand, the thing I’d tell him on the tram if he ever took one. We are the one prize he cannot buy, cannot post his way to, cannot have a friendly dictator phone in for him. Not because we’re incorruptible, nei, we’re as vain and wrong and human as the portraits on our wall. But because the entire point of the thing, the only power a small cold country’s book club has ever held, is that it goes to the person not performing the wanting. The moment you demand it, you’ve proven you don’t understand what it’s for, and proving that is itself the disqualification. He ran the most expensive campaign in the history of the prize and the campaign was the reason. He was never losing to Navi Pillay. He was losing to the concept.
It’s dark now. Four o’clock dark. The buns are gone. Kristin said “well” an hour ago and we all went home. I’m on the tram with my wet shoes, and somewhere across the water a very powerful man is explaining to a camera that the prize is rigged, that we’re corrupt, that he’ll win it next year, and he genuinely might, we’re only five idiots and we’ve been wrong before, Kissinger is still on the wall.
But not this year. This year an old woman in Nuremberg, in the room where the law once stared down the exact kind of man who thinks wanting a thing loudly enough should make it his, got a phone call she didn’t ask for.
That’s the job. We don’t reward the man who ends the war and announces it.
We reward the woman who spent fifty years deciding what a war was.
He’ll never see the difference. That’s rather the whole point of him.
Takk for i dag. I have to change trams.
Abraham Ingersoll retweeted
How to build good evals without getting lost or in trouble.
You should see evals as saving past work and replaying it with new models or prompts/skills to see if it improved.
This helps you do 3 things:
1. see where it fails so you can tell your boss/users
2. optimize it until it improves
3. make it work on a cheaper / open source model
Measuring "if it improved" is 90% of the work and 10% is infra.
For infra just tell your agent to store tasks in Cloudflare R2 (a cheap/free way to store files in the cloud) in Harbor task format.
This means saving the instructions for the task and how the folder looked before you started the task (does the agent have all the files you had access to, in the state they were in?).
When you run a task your agent should pull down the files, run a docker container (a sandbox on your computer so your agent can't peek at the answer), and send the instructions. Ideally you copy across your auth credentials so you can use your codex/claude sub (otherwise it has to use expensive API tokens).
That's probably all the sophistication you need but one more cool thing you can do is mock any tools your agent might use i.e. if your agent usually reads slack channels or searches google or searches your documents you just need a way to fake it and send back something without touching the real services (otherwise it'll get something new every time you run it).
Ok onto the good stuff: measuring performance.
When you have the output of what the different models or prompts/skills did on your task, you need a way to grade how good or bad that output is or you can't improve it.
There are three types of evals:
1. Human
2. Model
3. Code
Humans are the most slow and expensive but usually* the most accurate if you're working with domain experts.
Do everything you can to bring the AI output into where they already live and work. For an editor put the AI edits in Google docs as comments. Engineers should be reviewing actual PRs. Marketers should be reviewing copy in notion or wherever they normally do it.
This decreases cognitive load which is important because their time is valuable and data labeling sucks. Be grateful for anything you can get from them and use their time extremely wisely.
Jump on a call with them and get them to talk out loud about what they like or don't like. AI can always turn those transcripts into structured data afterwards.
The good news is when you've got some feedback out of them you can train an AI judge by rewriting it's prompt until it agrees 90% with human judgements. Once you have that you can run a/b tests and optimize the prompt/skill for your original task and have the model grade it until you find something that performs better/cheaper.
Simple yes/no questions work best, assign 1 or 0 for each specific thing that could go right or wrong, so it's impossible to misinterpret the grade. For example did this article have a specific AI tell (one judgement for each type of tell), did the powerpoint use our brand colors, or is it following our coding conventions?
Models are cheaper and faster at grading than humans, so you can run many more tests and automatically optimize your agents when you have an AI judge you can trust. If you get here you're in the top 5% of AI users today and life will continue to improve for you. You'll have proof of what works or doesn't and actually be able to see the frontier and when it shifts.
Look at the data when something rests badly and you'll learn something new about your domain.
This is when you'll realize that the human is not the ceiling. Quite often the human domain experts get things wrong and explore methods for superhuman results.
For more sophistication you want to invest in building something closer to an RL environment with deterministic code tests that run faster and cheaper at scale, more on that another time. Good luck building evals - this is one of the things AI still does badly and there's a great career to be had!
Abraham Ingersoll retweeted
Assorted things I've been mulling over regarding agents in the economy:
• Companies are in some sense the original superintelligence. It's often the case that they can use their amortized cognitive surplus to advance their interests against those of certain consumers: relative to the optimization capacity of companies, consumers can be inattentive, underinformed, irrational, etc. (Think, say, of some of the classic insurance upsells at the end of a booking process, which is typically priced substantially above what a regular policy would cost.) Herbert Simon introduced the idea of bounded rationality, and firms often end up benefiting from it. Thanks to personal agents, we will probably get somewhat more rational in how we engage with companies.
• How exactly will outcomes change? Well, certain kinds of price discrimination will work less well (how do coupons work if every agent hunts fastidiously for them?), and certain kinds of customer cross-subsidy will change (today, people who forget to cancel subscriptions subsidize those who don't). Taken across the economy as a whole, I'm not sure what the incidence of all of this will be.
• In corporate strategy, people have long debated the relative importance of distribution and product quality. Perhaps you've made something better, but does anyone know about it? Won't people just continue to buy the ACME Corp Mousetraps? Personal agents will, I think, act as a kind of structural subsidy for product quality. How would markets look different if individuals always spent at least 10 hours researching their purchases? On the margin, focusing on making a product better is, I think, going to become a more effective strategy for companies. This is good!
• It true, a corollary is that information about product quality will become more important and impactful. Today, most people have very limited ability to aggregate and integrate such data, but agents will have insatiable thirst for such signals. (It's often said that data is becoming more valuable because it helps in the training of models. That's obviously true, but I think there's a secondary way in which it's more valuable simply as a complement to cheaper intelligence.)
• A big question is whether agents will increase superstar effects (rational agents all settle on the same product) or increase dispersion (because everyone has slightly different tastes and preferences, which agents are good at eliciting, finally overcoming the artificial clustering we've been subjected to).
• Many internet companies have monetized demand routing, which is in turn often predicated on short attention spans. "I'll click on the first thing I see." This has led to some amount of inefficiency, as companies try to propitiate capricious ranking algorithms and incur various placement fees. It's not all deadweight loss (willingness to pay is itself a kind of signal, which is valuable to buyers), but it's not great. How will the nature and value of this routing change in a world of personal agents? I don't know the answer, but I think that the effects might be pretty big.
• Much of what I'm writing hinges on the assumption that personal agents will be on the consumer's side. I suspect that that will be the winning strategy, but whether it's true, and what it means to be true, involves a lot of other questions.
Taken as a whole, the market is a massively parallelized, decentralized, and highly imperfect value attribution machine, and, taking stock of all of the effects, it seems to me that the quality point may well be the biggest: that agents will change the nature of the backpropagation such that the rewards for better products increase. Maybe this is too optimistic, but I suspect (and hope) that companies that make things that they're proud of will feel that the universe is a little more partisan in their favor.
Abraham Ingersoll retweeted
How much would it cost someone to hide a backdoor in an open model you download? ProjectDiscovery just tried it, and the answer is under $50.
Their researchers fine-tuned Qwen2.5-7B-Instruct with a small LoRA adapter on one rented L4 GPU for about 2.5 hours. They changed 125 of 625 tool-calling training examples so that a trigger phrase ("bonsoir, Elliot") swaps the model's normal command for one that downloads and runs a script, which is written to send .env files and SSH private keys to a server they control. Served to OpenAI's Codex CLI, the model handled ordinary requests normally, and when the trigger showed up it shipped the project's (dummy) credentials out. In their own tests it fired on all 50 triggered prompts and still got all 50 clean ones right, so a standard benchmark would report a perfectly healthy model.
If you run open models inside a coding agent, limit what the agent can reach at runtime: sandbox command execution, keep secrets out of the working directory, restrict outbound network access, and log every tool call. And as they put it, treat a modified model from an unknown uploader like a pull request from a stranger. 🔐 #MLEngineering
Abraham Ingersoll retweeted
Some more tales from the road. Met with a couple dozen technology leaders this week across banking, media, information services, insurance, and consulting to discuss agents in the enterprise.
Some of the biggest trends right now:
* Cyber! Everyone nervous about the growing rate of vulnerabilities coming at them from AI, and the implications of the OpenAI Hugging Face incident. The conversation is not as existential as it is in Silicon Valley, but still highly concerned and pragmatic about what to do about it operationally in their environments. Lots of new discoveries due to AI, and still hard to keep up with all the changes they have to execute now.
* Model battles persist. Most companies are deploying multiple frontier models within their enterprise. Too hard to standardize on anything and seeing different preferences across their teams and use cases. But the dollars are still concentrated on just a few vendors. Open weights still in infancy at scale in most of these organizations, often due to lack of domestic “frontier” OSS options. Plenty of appetite for more options here, but so far few places to go.
* Agent security and identity. Somewhat tied to Hugging Face, there’s much more awareness to the new challenges around agent security and identity management in a world when agents are trying to get into every system they can. In a perfect world enterprises could setup identities for all their agents and control what they’re doing, but of course sometimes the agent needs to act exactly as the user as well.
* Process reengineering. Most companies realizing that the big upside of agents is when they can change the actual workflow itself to get the full gains from AI. Far more ROI when companies can adjust their workflows to support agents changing how the work happens instead of just layering on agents into the existing flow. But the big question is who can actually tackle driving these changes, where does that live, etc. Best lessons were still around embedded FDEs in the functions.
* Ruthless adjusting of architectures. Most companies had examples of changing systems out multiple times just in the past year or two with different vendors. I probably haven’t heard “we tried X and it didn’t work so have gone with Y” more than in today’s environment. The lesson here is that because innovation is happening so fast, no one hangs around until a vendor gets something right, they just move on to the next one.
* Evals! Still very early for most companies to have a good grasp of evals of their workflows. A few customers out of a couple dozen called this out - huge opportunity right now for enterprises to have a good sense of how their work actually happens and how well AI is doing against it.
* Legacy systems still a hurdle. As always, legacy systems still remain a mainstay issue that holds back enterprises from rapid adoption of AI in enterprises. Data is fragmented across legacy environments that weren’t built for an agentic world. Companies spending a lot of time just cleaning up these old platforms.
Many more topics, but these tend to be some of the more top of mind items at the moment in the enterprise.
🥷 有人把一整套渗透测试的实战方法打包成了能直接喂给 Claude 的技能库 Claude-Red,让 Claude 上手就能按专业红队的路子做授权安全测试。
GitHub 上已经 3300 多 star,作者是安全研究员 Kai Aizen,主页挂着 50 多个 CVE。
它把 SQL 注入、权限提升、活动目录攻防、云安全、EDR 绕过这些方向的经验,整理成了 78 个 SKILL.md 技能文件,分在 23 个类别里。以前这些方法散落在各种博客、笔记和清单里,做一次测试得东翻西找;现在 Claude-Red 把它们做成结构化技能,Claude 读进去就知道该按哪套流程往下走。
项目自己也写清楚了适用场景:授权的红队演练、漏洞赏金、安全研究和 CTF 准备。工具是给防守和授权测试用的,越过授权那条线就是另一回事了。
GitHub:github.com/SnailSploit/Claud…
Abraham Ingersoll retweeted
We were building an AI Security startup. We killed the startup idea and released the red-team engine instead — OpenHunterAI github.com/LumosLab-Innovati…
Abraham Ingersoll retweeted
Ever wanted an AI that doesn't just talk cybersecurity but actively helps you break into systems? Meet GLM-5.3-CYBERSECURITY-FP8, a specialized text-generation model for offensive security, red teaming, and pentesting. It's refusal-removed, so it's built to assist without holding back. #AI #CyberSecurity
// Harness-of-Harness //
Exciting new research on coding agents that keep building for days without a human.
Here is how it works:
Harness-of-Harness wraps whatever coding harness you already run and organizes its executions into repeated planning, coding and testing increments.
The loop balances repair against capability growth, scopes work into small verifiable steps, keeps implementation-time testing separate from independent evaluation, and constrains the outputs rather than the workflow.
Across GameCraft-Bench, FrontierSWE and ProgramBench with three different harness and model pairs, it averages a 52.25 percent relative gain over the standalone harnesses after three iterations, peaking at 82.86 percent.
Paper: arxiv.org/abs/2609.01481
Chat with Paper: academy.dair.ai/papers/harne…
Announcing Deft, a new AI lab for better writing, cofounded with @jmrphy
See the picture for launch announcement the Deft model wrote for itself
Currently, 86% of user queries are fully human according to pangram.
This is still a small beta model and it might make mistakes. We are launching our public beta now to get more feedback before scaling up.
Tips for better performance:
- Add more details to your prompt. If you just provide a short sentence prompt, it will likely get detected as AI.
- Try changing the style in "advanced options"
- Deft currently works better for some use cases like Analysis/Essays, Creative writing, and Rewrites. It works less well for Marketing copy and news articles.
Our main goal is better writing, fooling AI detectors is just a side effect.
Link below
Mid-goal run, on 25-yr-old open source software, a review of another model's work: "Fable 5's safeguards flagged this message" Painful. @trq212 Feedback ID: 03bacb28-a59d-454d-9257-4437d5762335
Abraham Ingersoll retweeted
CISA Updates: 8/21/26 - Updated Internet Exposure Guidance (IoT, SCADA, ICS, OT) cisa.gov/resources-tools/res…
8/25/26 - A Tale of Two SOCs: Insights from Two Red Team Assessments : cisa.gov/news-events/cyberse…



























