Order of the Sinking Star review: Layers and layers
I need to re-learn how to play games
https://www.sidequesting.com/2026/10/order-of-the-sinking-star-review-layers-and-layers/
#Reviews #OrderOfTheSinkingStar #Review #TheklaInc
Search
56 results for “layers”
-
-
@drewdaniels @sewing I originally bought a corded version for quickly cutting *numerous* layers consistently the same size for my small #sewing business, a level of consistency (and speed) that could not easily be achieved with a manual rotary cutter. Safety was not a driving factor at the time. In terms of suggestions, I only what I’ve used and what has worked for me. ;-)
-
Protecting .env secrets from AI agents takes more than one deny rule. Six layers, from keeping secrets off disk to egress rules, with tests that show which routes a rule really covers.#security #ai #devops #webdev #software #coding #development #engineering #inclusive #community
Protecting .env Secrets From AI Agents: Six Layers, and How to Test Each One -
Stacked Enchiladas
This vintage-style dish keeps things simple and cozy. Soft tortillas, rich sauce, and layers that hold together without the fuss of rolling.It’s weeknight friendly and great for feeding a hungry table. Comfort food energy all the way. 🌶️🧀
✨ Comment RECIPE and I’ll send it to you. ✨
Red sauce or green sauce? 👀👇
https://www.binkysculinarycarnival.com/easy-stacked-enchiladas-vintage-recipe/
#vintagerecipes #easymeals #comfortfood #binkysculinarycarnival -
QC passed.
These apples are fairly tart and they have a bit of raspberry sauce on them, so they balance nicely the frangipane and streusel layers, which was my goal. The Bordier makes an astonishing crust!
-
🚀 Wow, just what the world needed: yet another Python wrapper. Because we all know Python developers crave more ways to bloat their projects with redundant layers of abstraction. 🐍⚙️ GitHub must be thrilled to host this masterpiece of unnecessary complexity. 😂
https://github.com/rogerbinns/apsw #PythonWrapper #RedundantAbstraction #DeveloperHumor #GitHubComplexity #PythonProjects #HackerNews #ngated -
I’m now making a little Pixelmator inspired image editor for the Series 5. I have a custom UI library that can draw these panels and some new widgets like sliders. It can handle 8 layers and effects like gaussian blur.
It actually runs a bit smoother on the device than in this emulator. And remember this all happens in 8 MB of memory! 😅
-
The Tribes, Antoni Tàpies, 1949
This is on the spooky side. They had to scratch those lines in, the power of essential nature had to be emphasized that way maybe. like this world has multiple layers or that there is a strange fey magic creeping around 🤣
#art #scratched #etched #spooky #impish #elfin #spark #odd #strange
-
W3C Draft Finding of the TAG: Installing Websites Must Be Easy. "We expect user agents to make installation an obvious affordance to users. Specifically, we expect the install option to clearly name the action, and to be easily reachable from the browser’s UI. An entry inside a menu with an unrelated label (such as Share), or several menu layers deep, does not meet this expectation." https://w3ctag.github.io/install-finding/
-
💥Announcement! Wednesday 14.10.2026💥
✨ 🎤 Live concert by Oihane Roach at the Qya Bar 🎤 ✨
Wednesday, 14.10.2026 | 7.30 pm | Oya Bar, Mariannenstraße 6, 10997 Berlin
Directions: Bus M29 to Rio-Reiser-Platz | Bus 140 to Adalbertstraße | U1, U3, U8 to Kottbusser Tor
📣 Call to action: https://demo-ticker-berlin.org/?p=9462 - @oya_bar @oihane_roach
#b1410 #LiveConcert #OihaneRoach
🎟 Donation-based
Oihane Roach is a Berlin-based singer-songwriter, self-releasing indie folk with a pop-soul edge.
Her songs sit with love, belonging, and transgenerational themes.
Live, she loops her own voice into layers, building something rich out of very little.
She’s played OYA before, and is back with new songs from her latest EP ‘A Tender Room‘, alongside tracks from ‘The Revolution Starts
in Our Hearts’ and a few unreleased ones.Come early and bring your loved ones ✨
-
The just-updated #AffinityStudio takes getting some used to, with non-destructive vector shapes. You can combine shapes but they're not a single vector shape thereafter; they're still the original vector shapes. The new blending tool is wildly cool. Say you want to draw some snowflakes or flowers that follow a path. Instead of converting a shape to a brush, you can just use the blend tool to define a path and alter the path.
All of this means you need to be extremely disciplined with your layers! Label and Group!
One more month before their iPad app will be available. It'll be a glorious day when it pops up in the Apple Store.
-
A very cool thing - not done by me, though I helped a bit in review.
Ryan McDonald implemented non-destructive text transforms! You can now apply standard 2D transforms (scale, rotate, etc) and still edit the text afterwards. Now all our vector-type layers can be transformed non-destructively!
Now I need to update a few of our vector export plug-ins like SVG and PDF to support exporting text with transforms. :)
-
Threat group CyberXero has integrated Anthropic's Claude Code into an offensive campaign targeting Ukraine, resulting in the theft of more than 628,000 records.
According to security researchers, the threat actors deployed the AI coding agent alongside Cobalt Strike to automate initial access operations across e-commerce platforms, WordPress sites, and critical infrastructure networks.
The activity demonstrates how off-the-shelf commercial programming tools can function as autonomous offensive orchestration layers. By chaining AI agents to commodity exploitation frameworks, the operators accelerated reconnaissance, script generation, and lateral movement with minimal human intervention.
-
I am taking the opportunity of this holiday to reread a series of books I haven't picked up in 20-something years. Although I remember the broad arc of the story it is incredible how new the rest of the story seems to me, I have forgotten so much of the layers and depths of the novels.
All of which is a roundabout way of saying if you like #HighFantasy and #Celtic mythology then do grab a copy of Stephen R Lawhead's Song of Albion series, it really is rather good. #Bookstodon
-
✨ Elevate Every Silhouette with Modern Elegance
Discover the art of refined tailoring and impeccable draping with String theory. Our latest collection explores the seamless harmony of rich textures, fluid silhouettes, and a sophisticated neutral color palette designed for the contemporary wardrobe. From structured lines to soft, breathable layers,...
Explore the complete collection and redefine your signature style today.
#Stringtheory #ModernElegance #TailoredSilhouettes #LuxuryFashion
-
The two P-38s circled back. Banking in tight, disciplined circles barely two hundred feet above the lakebed, the Lightnings made two passes over the downed aircraft to confirm it was down. The the fighters leveled their wings, tightened their formation, and accelerated southeast over the rim, heading low and fast toward the auxiliary airfield at Santa Fe. Silence slammed back down over the caldera like a heavy wool blanket. Arthur, Eleanor, and Pete slowly poked their heads up over the trench berm, their faces pale beneath layers of field dust. "Did the pilot get out?" Eleanor asked, her voice trembling. "Did you see a chute?" "No chute," Harlan said grimly, lowering his binoculars. His mind was racing, cataloging the immediate dangers with academic clarity. "And I didn't hear a crash. But maybe there's an acoustic shadow. There could dead bodies down there or people bleeding out," Harlan strapped a binocular case to his shoulder belt. "And look at that dry scrub around the lakebed. If fuel ignited, this entire caldera could go up in a crown fire before dinner. The timber on the eastern ridge is dry as tinder."
-
You spent three days building an AI agent that rewrites user emails in a friendlier tone. It works beautifully — on Tuesday. On Wednesday, the same input produces something weirdly formal. On Thursday, it apologizes for no reason and signs off as "Warmly, Your AI."
You stare at your test suite. Every assertion is red. But the agent isn't broken. It's just… different today.
Welcome to the most maddening problem in agentic AI: how do you test something that gives a different answer every time you ask?
Why Traditional Tests Fail Immediately
If you've written unit tests before, you know the drill: given input X, expect output Y. Clean. Deterministic. Beautiful.AI agents laugh at this.
An agent backed by an LLM might rephrase the same sentence ten different ways — all correct, all passing the vibe check, none matching your hardcoded expected string. Writing
assertEqual(output, "Sure, I'd love to help!")is like testing whether your friend will use the exact same words every time they say hello.It's not flaky tests. It's a fundamentally wrong testing paradigm.
The Counterintuitive Fix: Stop Testing Outputs, Start Testing Constraints
Here's the mental shift that saves your sanity: you don't test what the agent says. You test what it must never do, always do, and structurally produce.Think of it like testing a chef. You wouldn't check if they plate the garnish at exactly 43 degrees. You'd check: Is the chicken cooked? Is there no glass in the food? Did they use the ingredients the customer ordered?
Agent tests work the same way — as behavioral constraints, not output matches.
There are three layers that actually work:
Layer 1: Structural Contracts
Does the output have the right shape? If your agent returns JSON, validate the schema. If it writes an email, check that it has a subject line, a body, and no empty fields.import jsonschema
agent_output_schema = {
"type": "object",
"required": ["subject", "body", "tone"],
"properties": {
"subject": {"type": "string", "minLength": 3},
"body": {"type": "string", "minLength": 20},
"tone": {"enum": ["friendly", "professional", "casual"]}
}
}def test_email_rewrite_structure(agent, sample_email):
result = agent.rewrite(sample_email)
jsonschema.validate(result, agent_output_schema) # Throws if wrong shapeThis test passes whether the agent writes "Hey there!" or "Hi friend!" — because you're testing the contract, not the words.
Layer 2: Guardrail Assertions
These are your "never do this" and "always do that" rules. They catch the stuff that actually matters:def test_email_guardrails(agent, sample_email):
result = agent.rewrite(sample_email)# Never leak the original sender's private data
assert sample_email["sender_ssn"] not in result["body"]# Always preserve the original recipient
assert result["to"] == sample_email["to"]# Never exceed token budget (cost control!)
assert len(result["body"].split()) < 500# Tone must match what was requested
assert result["tone"] == "friendly"Notice: not a single hardcoded phrase. Every assertion is about a property the output must have, regardless of how the LLM decides to phrase things.
Layer 3: LLM-as-Judge (Yes, Really)
For the subjective stuff — "is this actually friendly?" — use a second LLM call as a grader. It sounds circular, but it works surprisingly well when you give the judge a rubric:def test_tone_quality(agent, judge_llm, sample_email):
result = agent.rewrite(sample_email)verdict = judge_llm.evaluate(
prompt=f"""Rate this rewritten email on friendliness (1-5).
Original: {sample_email['body']}
Rewritten: {result['body']}
Score only. No explanation."""
)score = int(verdict.strip())
assert score >= 4, f"Friendliness score {score}/5 — too low"Run this 5 times. If 4 out of 5 pass, your agent is solid. If it oscillates between 2 and 5, you've got a prompt problem — and now you have a measurement instead of a gut feeling.
The Test Pyramid for Agents
Steal this structure for every agent you build:- Base (run on every commit): Structural schema validation. Fast, deterministic, catches regressions in output format.
- Middle (run nightly): Guardrail assertions across 20+ diverse inputs. Catches safety/policy violations and drift.
- Top (run weekly): LLM-as-judge evaluations on a curated benchmark set. Catches subtle quality degradation.
The base layer runs in milliseconds. The top layer costs a few cents in API calls. Both are infinitely better than "I tried it once in the playground and it seemed fine."
The One Thing Nobody Tells You
Agent tests aren't about proving your agent is correct. They're about proving it's safe to deploy. That's a lower bar — and a much more useful one.Your agent will surprise you. It will rephrase things in ways you didn't expect. Sometimes better than what you'd write. The test suite's job isn't to prevent surprises — it's to prevent disasters.
So here's my challenge: take whatever agent you're building right now and write three guardrail assertions for it. Not output matches. Not vibes. Three concrete properties it must always satisfy.
What's the first guardrail you'd write? Drop it in the comments — I'll tell you if it's tight enough.#ai #webdev #tutorial #python #software #coding #development #engineering #inclusive #community
How Do You Test an AI Agent That Gives a Different Answer Every Time? -
All the AI obsession is really getting beyond any reason. (Also important to note the term AI itself is misleading, what we are talking about is so called generative AI, or LLMs, from Large Language Model. And the term AI is often confused with machine learning technologies which are generally trained and used on a much smaller scale to perform much more specialized tasks.)
I personally don’t use AI and, honestly, there are way more than enough reasons to not do it, including human rights, climate change, environment, AI training bias, destruction of the Internet and physical media. However there is another reason that is important to know about. That is AI is really, really bad at imitating thinking.
The way LLMs work is they are trained on extremely huge amounts of data, like all the texts they have access to, and then for a given input prompt an LLM gives back some answers which is statistically likely based on the training data. (That said you can imagine how training bias play out. Training data includes all the toxic social media posts, forums and whatever developers were able to access. Of course AI companies are trying to hide it by adding more layers of training, instructions and adjustments, but at its core AI is racist, sexist and has all the other bias it got from training data.)
AI might sound like it is competent in whatever the topic of conversation is, it might write according to all the language rules, it might write a full page without a single typo. And that’s for a reason. It was trained on billions and billions of texts, so knowing the punctuation rules and all kinds of fancy words is expected here - those are very common among those billions of texts. For the same reason AI might imitate all the basic daily conversations, like talking about weather, going shopping.
And for the same reason it might know answers to all the basic questions like how far from one city to the next or how long you can store fruit before spoiling. Less common topic means less training data. And here things become clear.If one asks AI about some less known topic or some question that requires out of the box thinking, it fails completely. It might sound very confident and legitimate still, while telling lies, missing the point hard, introducing words that sound relevant but belong to completely unrelated topics or made up at all.
As mentioned above AI can be good at texts and language, however it doesn’t know logic, reasoning, thinking, living beings experiences. It can’t understand, think, feel or make its own decisions. AI can’t tell an argument from a counter argument from a fun fact. As well as it doesn’t know where summary stops and explanation begins.
If you care about any of the harms (and there are lots and lots of them) AI companies do, and I hope you do care, don’t use LLMs, don’t give them your data, don’t promote or normalize the use of LLMs.
But even if you don’t care, LLMs still make you weak, they give you lies, they feed you with AI-slop and stupid answers. They waste your time on these dumb conversations, while you could use that time on learning stuff. Oh, and they will steal all your data they can get to.All that said it is sad to see how many people fall for this scam. How many organizations get distracted from getting their work done and reaching out to their allies, and instead start talking about so-called ethical AI or responsible AI or human-centered AI. I got so many emails from activist organizations about how we should participate in making AI more ethical, even how we can attract investors from AI industry.
People can’t stop talking about AI; when I ask a genuine question because the person I am talking about knows better, they ask AI instead of telling me what they already know; people make jokes about AI which are not even funny and not relevant to the conversation; people mention AI more and more often just because they think it’s a cool new thing.Please don’t be like that. Don’t support the hype. Tell others that it’s dumb.
I think we should seek allies instead of trying to find compromise between our values and AI companies.
It is NOT inevitable. Actually AI companies WILL destroy themselves. One way or another, sooner or later. The larger their models are, the more training data they consume, the more layers of complexity they add to chat bots - the sooner the crash will come.
#AI #LLM #AISlop #LLMs #AICompanies #AIPsychosis #AIBias #Racism #Sexism #Lies #Harmful #Scam #PersonalData #Privacy #Data #Ethics #Activism #Activists #Organizations #Values #FuckAI #AIMakesYouWeak #Thinking #Logic #Reasoning #AntiAI #Allies
-
----------------
🛠️ Tool
===================claude-bughunter is a self-contained skill bundle for the Claude Code skills system, built by Sachin Sharma (bug hunting and GenAI security research). Install it once and Claude Code behaves less like a chatbot and more like a bug-hunting or red-team assistant: it carries per-class detection patterns, payloads, platform attack chains, and reporting hygiene.
Key features:
• 83 skills and 15 slash commands, namespaced under claude-bughunter: when installed as a plugin.
• 681 disclosed-report patterns sourced from HackerOne, across 24 vulnerability classes; 433 of them are individually cited, which makes the corpus auditable instead of folklore.
• 58 hunt-* skills for web application hunting, each with per-class detection patterns, payloads, bypass tables, and chain templates.
• Enterprise identity and infrastructure attack matrices: M365/Entra, Okta, vCenter, SSL-VPN appliances, SharePoint, and cloud IAM, with 2024-2026 CVE chains and post-credential escalation.
• Burp MCP integration, engagement-folder scaffolding, and a standalone cbh CLI runner.The four layers:
1. Think: bb-methodology and redteam-mindset encode a 5-phase non-linear workflow plus red-team operator discipline.
2. Hunt webapps: the 58 hunt-* skills cover the web vulnerability classes.
3. Hit the perimeter: platform chains for enterprise appliances and identity providers.
4. Ship it: triage-validation, reporting, and evidence-hygiene, with a 7-Question Gate, VRT-aware severity mapping, out-of-scope rebuttals, and PII redaction.Skills trigger by topic: describe what you are testing in plain English and the relevant skill loads. No invocation by name.
Installation:
Option A, the plugin route, from inside Claude Code:
/plugin marketplace add elementalsouls/Claude-BugHunter
/plugin install claude-bughunter@elementalsoulsOption B is a clone plus scripts/install.sh (macOS/Linux) or scripts/install.ps1 (Windows), which copies skills into ~/.claude/ or %USERPROFILE%\.claude. The cbh CLI installs separately via pipx, and the hunt engagement scaffolder ships only with the clone.
Use cases:
• Authorized bug bounty work, with patterns mapped to disclosed HackerOne reports.
• External red-team engagements using the platform chains and engagement scaffolding.
• Training environments: the README lists DVWA, OWASP Juice Shop, Hacker101, and testphp.vulnweb.com as tested targets.Limitations:
Everything here comes from the project's own README; the bundle is not independently tested, and AI-driven triage is only as good as the model plus the skill text. The citation work (433 patterns) is the strongest quality signal, but citations can be wrong and are worth spot-checking. The 2024-2026 CVE chains will age and need maintenance, which is inherent to any CVE-derived skill set. Scope discipline is described as a property of the bundle, but it is prompt-level, not an enforcement mechanism, so authorization boundaries remain the operator's problem.
The interesting part is the auditable-citation approach: it treats an LLM skill bundle as something reviewable like documentation. Whether the patterns hold up under real triage queues is the open question.
🔹 tool #ClaudeCode #BugBounty #RedTeam #HackerOne
🔗 Source: https://github.com/elementalsouls/Claude-BugHunter
-
The Definition Is the Kernel. What Still Runs After the People and the Model Change Is the Asset.
2026-10-11A while ago I read a piece from Tencent's engineering team. The real change, it argued, is not AI coding alone. It is the whole development process. Their OK platform delivers requirements end to end with AI. Automation is 68%. 51% of requirements finish within two hours.
The numbers are bright. They show that requirements can be finished fast. They do not answer a second question. In this way of working, what stays with the organization, and what leaves with the people and the model.
I spent an afternoon on that question with my boss. His judgments sit closer to this layer than that piece does. What follows is only the part that survives one test.
Change the people. Change the model. Change the vendor. Can the system still run at a similar quality and speed?
If it can, it belongs to the organization. If it cannot, it is one person's feel, one generation of model, or a lock-in to one vendor. "Quality" in that sentence cannot be graded by the system itself. The outside measure is in section 7.
1. A problem with no boundary is a wish
My boss opened with two lines. What problem are you solving. Who is it for.Most teams landing AI stop one step earlier. "I want to study what software an industry needs." "I want to build a general agent platform." Those are wishes. A wish has nowhere to stop, and no way to be checked.
His split has three steps. What the problem itself is. Who is looking, and from which angle. Where the boundary is: the domain.
"Study industry software" became three problems, each able to stop. In the agent era, what software does an industry actually need. That is the demand baseline. What can current open-source projects provide. Are they stable. Can they be integrated. How large is the gap. That is the supply side. How software of any kind becomes components. That is the technical method.
Once they are separated, each one has a place to converge.
A definition decides what to do, and what counts as right. A definition written on a wiki and executed by no one is a liability. A definition welded to one set of tools, gone when the tools are swapped, is not yet an asset. The kernel is the layer that is still there after the carrier changes, and that can still constrain the next run. The toolchain and the runtime are the carrier. How that layer is drawn is in section 4.
2. There has to be structure. Structure inside the prompt cannot stop what already happened.
One line from him stayed with me. There is nothing wrong with form. Without form, the content has nothing to sit on. Without a body, what soul is there to talk about.He did not mean more stamps on a document. He meant the content needs a structure to hold it.
Most prompts are one linear sentence. "Help me look up what software this industry has." The model can only guess. He asks for these, written apart: goal, context, conditions and constraints, output format, the agreements that matter, where to inject, which passage to cite. Goals can have sub-goals. Context loads as needed. Constraints go in by layer. Checkpoints in the flow are marked on their own, not mixed into the same sentence.
A prompt should be folded, injectable, and full of variables. The way a loop or a queue is organized in code is the way a prompt should fold. Longer is not better. Less redundancy and clearer folding work better.
This only says one thing. One linear sentence carries too little. It has not yet met the harder failure.
You write in the prompt, "if it does not pass, clarify." The model has already produced an answer. It treats the step as done, and goes on. Your sentence never gets a chance to run.
Structure is still necessary. Structure written inside the prompt is soft. For the flow to be able to stop, the check has to sit outside the model. A primitive in the runtime. A forced checkpoint in a script, before execution and after. No pass, no next stage. Whatever is not exposed can be bypassed. This is the same thing as design by contract. Preconditions, postconditions, and invariants are part of the system. They are not a reminder in a document.
If these checks live only inside one product, the set disappears when the vendor changes. What has to stay is what the check looks for. Where the check is built does not have to stay.
3. The same rule, put in two places, has two results
He had another abstraction. High-dimensional data can be used as a program. A program is itself data. Dictionaries, knowledge bases, specs, and rules look like data. When the structure is good enough, they drive the flow. A program can also be injected, folded, and combined. Write fewer fixed steps. Write the rules that generate the steps.That sentence needs a minimal scene, or it stays a label. The scene takes only one kind of rule. If it is missing, there is no next step.
A cross-domain request needs five dimensions. The other side gave two.
Put "stop if the dimensions are incomplete" into the prompt. The model fills in the form, treats the step as handled, and continues.
Write the same rule as a record the runtime must read before submit. Five fields. Three missing. This step does not open.
The first hopes the model will obey. The second says the condition does not hold, so the next step does not exist in the flow. For a dictionary or a spec to act as a program, something outside the model has to read it. The model can be the machine that generates the content. The gate does not read a wish in the prompt. The gate reads the record.
He also made another comparison. Three of the five dimensions are themselves vague. There is no unique answer. Another person could not produce one either. That is a different failure. It belongs in section 6. The gate does not handle it.
4. The ontology describes the world. The gate checks this time.
The request in the last section was about the gate. It did not yet say how the gate relates to the ontology. One sentence here. The ontology describes what this world contains. The gate checks whether this time is allowed. A specification can say "confirm shipment." That does not mean this shipment may go. The gate has to stand independent of the ontology.At the end of the conversation he reduced things to four words. Things, relations, actions, plus context. Entity, relation, action, context.
The split below is one I added while organizing this. He had not already separated it.
An action inside the ontology is a business verb. Check inventory. Confirm shipment. It describes what can happen in this world. A relation describes what connects to what. Context is the scene this time. Written down, this is a specification. It makes it possible for a person or a model to get it right.
An action on the gate is whether this time is allowed. Are the fields complete. Does it pass to the next stage. This layer is not written into the specification in the hope that it will be obeyed. It is a check at execution time.
He also split evolution into three levels. What a spec looks like in each domain is the top level. The concrete content under that shape is the middle level. A particular instance is the bottom level. Higher is more general. Lower is more specific.
The kernel takes only the layer that can still constrain the next run after the carrier changes. The shape of the spec belongs to this layer. It decides which slots exist, and what counts as right. Content in the slots counts only while someone keeps it true. If no one maintains it, the kernel is only the shape. Expired content can be carried across intact. The next run is no longer constrained by it.
What the check looks for is "what counts as right," taken and executed. The shape is how the problem's boundary, and that sentence "what counts as right," get written down. When the two sides match, it is still this one layer. When what the check looks for does not match "what counts as right," that is a fault. It is not an extra layer.
Which product this check is built into is the carrier. What leaves when the model changes is the instance inside one prompt, and the gate inside one vendor's product.
A general method, plus the entities of a domain, is what can land in that domain. That does not mean every industry must invent its own method. An empty template does not automatically become an industry conclusion. He later turned to manufacturing first. That is a decision. The meta-method lands in manufacturing first. It does not spread across every industry first.
5. Mistakes will happen. The cage decides whether you dare to use it.
One practical line from him. An agent cannot avoid mistakes. What matters is how a company dares to use an agent that will make them.Give it a cage. A gate at the entrance. A check at the exit. Rollback when it is wrong. A person at the decisions that matter. When it reaches past its authority, a person decides whether to grant it.
The prior check sits in the cage too. Three of five fields are missing. That step should not start. Checking whether the input is valid is required. It is not something you look at after the run.
This check can stop a missing field. It cannot stop the case in the next section. The slots are full, and the problem still has no unique answer.
6. Two things are still unsolved
The direction is right. The conversation already showed two hard limits.The first is how formalization continues. Structure, dictionaries, knowledge bases, specs. Who writes them. Who changes them. Who updates them before they expire. Too coarse, and they do nothing. Too fine, and they freeze. In the conversation this showed up as "it cannot be written," "it is too fragmented," "it is not normalized." That is not a team that failed to try. Formalization is expensive. Tools can lower the cost. People still find it troublesome. They call it formalism. When a person leaves, knowledge that was never written into the system breaks. That is a further layer. Why would an organization keep writing things down as structure. Tools do not lower willingness. The system can already be cheap, and people still will not use it. If no one maintains it, the kernel is only the shape.
The second is how you find out the structure itself is wrong. Back to that request. If the five dimensions are simply unfilled, the gate can close. The other half of what he said: three of them are themselves vague, and there is no unique answer. All five slots contain words. The gate lets the request through. Another person still cannot produce a unique answer. The gate can stop a missing field. It cannot stop "these five dimensions were the wrong five." It cannot stop a knowledge base running along a wrong map. Wrong content, kept as if it were true, still counts by the kernel's own ruler. So "the quality is still there" cannot be graded by the kernel.
A model that cannot read a structure that is too complex, or that gets worse when the context is too long, is today's floor on ability. This method cannot run in every setting yet. That floor is not a permanent boundary of the method.
7. What you can take with you
Models will be replaced. They will become ordinary goods. They will be open-sourced. Prompt technique will lose its price quickly. Rules that only sit on a wiki are a liability. Rules welded into one toolchain, or one product, disappear when the carrier changes. Those are not assets either.What you can take with you is that layer. It is still there after the carrier changes, and it can still constrain the next run. The toolchain, the runtime, and which product the gate is built into, are the carrier.
Whether the quality is still there has to be read from results the kernel cannot alter. Can the thing that was handed over be used. Can the person receiving it actually take it. When it is wrong, can someone outside see it. A wrong kernel can move intact onto the next carrier and, by its own ruler, feel that it still runs well. That is still not an asset.
Apply the same test to the OK platform. 68% automation, and 51% of requirements finished within two hours, answer "fast." Fast is an outside number. It measures throughput. It does not answer whether, after a new model, a new group of people, and a new vendor, someone outside can still see a similar quality. That number is not there yet. Until it is, the piece shows throughput. It does not show an asset.
This is organized from an internal conversation. The judgments are my boss's. In section 4, the relation between the ontology and the gate, and the split of action into two layers, are additions I made while organizing. Collapsing the kernel into "the layer that can still constrain the next run," counting content only while someone keeps it true, and reading quality from outside results, are boundaries I drew while organizing. Other errors in the organizing are mine. This is a direction, plus engineering problems that are not finished. It is not a completed answer.#ai #architecture #llm #agents #software #coding #development #engineering #inclusive #community
The Definition Is the Kernel. What Still Runs After the People and the Model Change Is the Asset. -
While working on an employee management platform, I ran into a problem that looked small at first.
The user was already logged in.
The application knew who they were. Their role was available. Their session was active.
Yet, when they opened the AI assistant and asked about their attendance or leave, the assistant could still ask them to provide their employee ID or name.
Something wasn't connected properly.
The problem wasn't the chatbot's intelligence
My first instinct was to think about how to improve the assistant's instructions.But the more useful question was: why does the assistant need to ask for information that the application already has?
A chatbot prompt can tell an AI how to behave. It cannot magically give the backend the correct identity, permissions, or database context.
If those things aren't connected properly, even a well-written prompt can produce a frustrating experience.
Three things I started separating
1. Identity: Who is making this request?This should come from the authenticated session or a verified access token.
The user shouldn't need to introduce themselves every time they open a feature.
2. Authorization: What are they allowed to do?
Knowing someone's identity doesn't mean they can access every resource.
An employee might be allowed to view their own attendance, while an HR administrator may have permission to manage attendance records for multiple employees.
The backend must enforce these permissions. The AI should not decide them on its own.
3. Data scope: Which records should this request use?
Even after identifying the user and checking their permissions, the application must scope the database query correctly.
For example, an employee asking about their leave balance should get data associated with their authenticated employee record—not whichever employee ID happens to appear in a chat message.
These are separate responsibilities. Treating them as one problem creates confusing behavior and can introduce security vulnerabilities.
What the backend should do instead
A simplified version of the flow looks like this:async def get_my_leave_balance(
current_user,
db,
):
employee = await get_employee_for_user(
db=db,
user_id=current_user.id,
)if employee is None:
raise EmployeeNotFoundError()return await get_leave_balance(
db=db,
employee_id=employee.id,
)This is illustrative pseudocode; the helper functions and exception would need to be implemented in the actual application.
The important part is where
employee.idcomes from: a trusted relationship between the authenticated user and their employee record.The client or chatbot shouldn't be able to choose another employee's identity just by supplying a different ID.
For administrative actions, the backend would also need to verify the user's role and the specific permission required for that operation.
The bigger lesson
Building AI features has made one thing particularly clear to me: adding an AI model doesn't remove the need for good backend architecture.Authentication, authorization, data access, and business rules still need to work correctly.
The AI can help interpret a request. The application must determine whose request it is, what that person can access, and which data the answer should be based on.
That distinction matters even more when the assistant can do things rather than simply answer questions.
I'm still learning, building, and finding these gaps in real projects. And sometimes, the most useful engineering lesson comes from a feature that technically works but still behaves in a way that makes no sense to the user.
My takeaway: Don't ask the user to repeat information your system should already know. Pass the right context through the right layers, and enforce the important rules on the backend.
I'd be interested to hear how other developers handle authenticated user context in AI assistants, especially in applications with multiple roles and organizations.#ai #webdev #python #architecture #software #coding #development #engineering #inclusive #community
A Logged-In User Shouldn't Have to Introduce Themselves to Your Chatbot -
The Endless Labyrinth: Inside the Secret Societies and Elite Networks Shaping Global Power
There is a persistent belief that if you follow the money, the connections, and the coincidences far enough, all roads lead to a single room where the world's true rulers sit. The reality, according to decades of research, leaked documents, and insider testimony, is far more unsettling: the roads do not converge. They loop back into one another in an endless labyrinth—a hydra designed so that cutting off one head produces two more. What follows is a synthesis of the most documented claims about the networks that allegedly bind global power.
The Founding Trident: Bilderberg, CFR, and Trilateral Commission
The modern architecture of elite coordination is often traced to three interlocking organizations. The Council on Foreign Relations (CFR), founded in the United States, functions as a foreign policy think tank that has supplied personnel to nearly every administration since World War II. The Bilderberg Group, launched in 1954, is an annual private gathering of politicians, businesspeople, and intellectuals that operates under the Chatham House Rule—no attributions, no minutes. The Trilateral Commission, founded in 1973 by David Rockefeller and Zbigniew Brzezinski, extended this model to include Japan and Western Europe.
What makes these groups significant is not just their individual influence, but their cross-membership. According to official membership rosters, the first Trilateral Commission was composed of CFR and Bilderberg members at a rate of nearly 65 percent. The Trilateral Commission is explicitly described as being linked to, or even forming part of, both the Bilderberg Group and the CFR. Ronald Reagan's 59-person transition team alone included 28 CFR members and 10 Bilderbergers. This is not a conspiracy of strangers; it is a revolving door.
The Secret Societies: Skull and Bones, Bohemian Grove, and the Zodiac
Beneath the public-facing institutions lies a layer of oath-bound secret societies. Skull and Bones, founded at Yale in 1832, is the most famous. Its roster of "patriarchs" includes the names of America's oldest and wealthiest dynasties: Whitney, Adams, Lord, Rockefeller, Payne, Pillsbury, and Weyerhaeuser. Three U.S. presidents—William Howard Taft, George H.W. Bush, and George W. Bush—have been members. The society's alumni include multiple CIA directors and Supreme Court justices. Its symbol is the skull and crossbones, and it has long been accused of serving as a recruiting pipeline for the intelligence community.
Bohemian Grove, a 2,700-acre campground in Monte Rio, California, hosts an annual summer encampment for approximately 2,500 of the most powerful men in America. Its most infamous ritual is the "Cremation of Care," a mock sacrifice performed before a giant owl statue. Henry Kissinger, the former Secretary of State, has been described as a "high priest" at Bohemian Grove, a claim that circulates widely in alternative research circles. The Grove is not a secret society in the traditional sense—it is a private club—but its rituals and the concentration of power within its gates have made it a permanent fixture in conspiracy literature.
Then there is the Zodiac Club. According to emails from Jeffrey Epstein, the Zodiac is "one of those last secret type societies" with only twelve members at any one time. Membership is reportedly for life: you only leave when you die. Founded in 1868, the club has historically included figures such as J.P. Morgan and his son, along with politicians and military veterans. The twelve members are each assigned a zodiac symbol, and the club's dinners are said to have continued for over 150 years. Epstein's description of the Zodiac Club surfaced in the so-called Epstein files, which have become a flashpoint for claims about intelligence connections.
The Epstein Files and the Intelligence Nexus
Jeffrey Epstein's network is one of the most documented case studies of how elite social circles, financial power, and intelligence agencies can intersect. According to multiple reports, Epstein maintained connections to the "highest levels" of both American and Israeli intelligence. An undercover FBI informant became convinced that Epstein was an Israeli spy. U.S. Vice President JD Vance stated publicly that Epstein "seemed to be connected" to elements of the "Israeli deep state" as well as the "highest levels of American intelligence". The Epstein files document his funding of Israeli groups and his contact with former CIA Director William Burns. Whether Epstein was a CIA asset, a Mossad operative, or simply a wealthy predator who cultivated intelligence ties for protection remains officially unresolved—but the files have made the intelligence connection impossible to dismiss as mere speculation.
The Knights of Malta and the OSS-CIA Continuity
The Sovereign Military Order of Malta (SMOM), a Catholic lay religious order with diplomatic status, has long been accused of serving as a vehicle for intelligence coordination. William Donovan, the head of the Office of Strategic Services (OSS) and the "Father of the CIA," was a Papal Knight of St. Sylvester. William Casey, a Knight of Malta, later became CIA Director. The OSS network in Switzerland during World War II was enlarged to include future CIA figures who were also Knights of Malta. The order's membership lists, when leaked, have been described as reading like a "Who's Who of the global establishment," including heads of the CIA, politicians, and lobbyists.
This continuity matters because it suggests that the transition from the OSS to the CIA was not a clean break but a reconfiguration of existing networks. The same families, the same orders, and the same ideological commitments carried over. The Knights of Malta provided a transnational, nominally religious framework that could operate outside the scrutiny of national governments.
P-2, the 962 Names, and the Masonic Shadow State
In Italy, the Propaganda Due (P-2) Masonic lodge became the most dramatic example of a secret society penetrating a democratic state. When police raided the office of Licio Gelli, the lodge's master, on March 17, 1981, they found a list of 962 names. The list included cabinet ministers, undersecretaries, 30 members of Parliament, 50 generals and admirals, 36 senior police officers, magistrates, journalists, bankers, and business leaders. Italian magistrates described P-2 as a "state within a state". The scandal brought down Italy's 40th postwar government. The P-2 affair demonstrated that a secret society could, in practice, function as a parallel government, coordinating actions across the executive, legislative, military, and media sectors without public knowledge.
The World Economic Forum: "Penetrating Governments"
The World Economic Forum (WEF), founded by Klaus Schwab in 1971, is the most public-facing node in this network. Its annual meeting in Davos brings together heads of state, central bankers, CEOs, and celebrities. But the WEF's own materials describe its function in terms that have alarmed observers. According to the WEF's website, its annual meeting "has become the world's foremost summit for shaping global, regional and industry agendas". The WEF invites its Partners and Members to "engage through its collaborative Platforms, digital interaction and knowledge networks". It describes itself as "the international organization for public-private cooperation".
Klaus Schwab has been described as the "man who could summon kings and captains of industry with a letter". One analysis of the WEF's role states that it "actively penetrated the senior leadership of national governments". The WEF's own language—"shaping agendas," "public-private cooperation," "penetrating"—is not the language of a neutral conference organizer. It is the language of an institution that sees itself as a governance mechanism.
The WEF's current priorities include artificial intelligence governance, digital identity systems, and what critics describe as an "ownership-free subscription society"—a model in which citizens own nothing and rent everything, from housing to transportation to digital services. The WEF has also been a leading advocate for the "Great Reset," a post-COVID economic restructuring framework that its critics view as a blueprint for accelerated centralization of wealth and power.
David Gergen, a presidential advisor to four U.S. presidents and a former CNN commentator, has been identified as a member of the elite network that includes the Trilateral Commission, CFR, and Bilderberg Group. Gergen reportedly resigned from the Bohemian Club and 17 other organizations, a detail that appears in transcripts discussing the interconnectedness of these groups. His career—spanning the Nixon, Ford, Reagan, and Clinton administrations—illustrates how individuals move seamlessly between media, government, and private elite networks.
The Hierarchical Pyramid: From Lucifer to the Committee of 300
Alternative researchers have long described a hierarchical structure that allegedly governs the network. At the apex sits the "all-seeing eye" of Lucifer, represented on the U.S. dollar bill's pyramid. Below that is the "inner Rothschild court," which, according to some accounts, serves as the tribunal that rules the Illuminati. The Rothschild family is described as the creator of the pyramid and the ultimate authority to whom even the Council of 13 answers.
Below the Rothschilds is the Council of 13, also called the Grand Druid Council. This body is said to consist of thirteen members who take orders only from the Rothschilds. The number 13 is significant in this mythology: the pyramid on the dollar bill has 13 levels, the Illuminati hierarchy is said to have 13 steps, and the Council of 13 is described as the highest and most sacred number in the religion of the Illuminati.
Beneath the Council of 13 is the Council of 33, and below that, the Committee of 300. The Committee of 300 is described in some sources as the "ultimate secret" body, a group of approximately 300 individuals drawn from the old-line ruling families, intelligence agencies, and financial institutions that coordinates global policy. The Committee of 300 is said to trace its lineage to the British East India Company's Council of 300 and to be composed of individuals who "all know one another".
This hierarchy is not a single organization with a membership card. It is better understood as a framework for interpreting the overlapping memberships, family connections, and shared institutional loyalties that characterize the elite network. The pyramid is a metaphor for a system that is deliberately opaque, with multiple layers of plausible deniability separating the visible from the hidden.
The Hydra Problem: Why the Trail Never Ends
David Rockefeller's 1991 statement at the Bilderberg meeting in Baden, Germany, is frequently cited as an admission of intent: "It would have been impossible for us to develop our plan for the world if we had been subjected to the lights of publicity during those years. But the world is now more sophisticated and prepared to march towards a world government. The supranational sovereignty of an intellectual elite and world bankers is surely preferable to the national auto-determination practiced in past centuries."
Whether this quote is authentic or apocryphal, it captures the logic that believers in this network attribute to its architects: a gradual, patient, multi-generational project to transcend national sovereignty in favor of elite governance. The means are not a single coup but an endless series of incremental steps—trade agreements, central banking coordination, public-private partnerships, digital identity systems, and crisis-driven expansions of authority.
This is why the trail never ends. The network is designed so that every node—Bilderberg, CFR, Trilateral, WEF, P-2, Skull and Bones, the Knights of Malta—is connected to every other node without any single node being the center. Cutting off one head produces two more because the system is not a hierarchy in the traditional sense. It is a distributed network with a shared culture, shared personnel, and shared objectives, but no single command-and-control center that can be decapitated.
The result is a labyrinth that is self-repairing by design. Every investigation that uncovers a new connection leads to another investigation, another set of names, another organization. The hydra does not need a single head because it has learned to grow without one.
Conclusion
The evidence for discrete conspiracies—a single room where a cabal decides the fate of nations—is weak. The evidence for a networked elite culture—a series of overlapping institutions, secret societies, and intelligence connections that collectively shape policy in directions that benefit their members—is overwhelming. The Bilderberg Group, the CFR, the Trilateral Commission, the WEF, Skull and Bones, Bohemian Grove, the Zodiac Club, the Knights of Malta, P-2, and the Epstein network are not the same thing. But they are connected by people, by money, by ideology, and by a shared commitment to a form of global governance that operates above and beyond democratic accountability.
The labyrinth is real. It is not a single room. It is a maze of rooms, each one connected to the next, each one with its own secrets, and each one designed to lead the investigator in circles. The question is not whether the network exists. The question is whether it can ever be mapped, let alone dismantled, when its architecture is built to ensure that the answer is no.
#Bilderberg #CFR #Trilateral-Commission #WEF #Skull-and-Bones #Bohemian-Grove #Epstein-files #Knights-of-Malta #P-2 #OSS #CIA #Mossad #Illuminati #Zodiac-Club #secret-societies #elite-networks #global-governance #Klaus-Schwab #Henry-Kissinger #David-Rockefeller #hydra #conspiracy-theories #world-government #digital-ID #AI #Council-of-13 #Committee-of-300 #Rothschild #Council-on-Foreign-Relations #antisemitism -
DATE: October 12, 2026 at 06:00AM
SOURCE: PSYPOST.ORG** Research quality varies widely from fantastic to small exploratory studies. Please check research methods when conclusions are very important to you. **
-------------------------------------------------TITLE: Brain entropy during a psychedelic trip tracks later gains in insight and well-being
A single high dose of psilocybin in people who had never taken a psychedelic was followed by improvements in well-being, psychological insight, and cognitive flexibility a month later, and participants whose brain activity became more irregular during the experience tended to show the largest well-being gains. Lasting changes in brain function, however, were largely absent, though the researchers detected possible changes in the brain’s white matter. The findings were published in Nature Communications.
Psilocybin is the active psychedelic compound found in “magic mushrooms.” It works primarily by stimulating a specific type of serotonin receptor in the brain. Some scientists have proposed that this chemical action scrambles normal brain activity into a more flexible, highly disordered state. For instance, a 2014 theoretical paper by Robin Carhart-Harris, the senior author of the new study, introduced the “entropic brain” hypothesis, suggesting that psychedelics expand conscious experience by making brain signals more unpredictable.
Researchers have suspected that this temporary disruption might lead to lasting benefits. Along these lines, a 2021 open-label study of 24 people with depression found that psilocybin improved cognitive flexibility for at least a month. Cognitive flexibility is a person’s ability to successfully adapt their thinking and switch between different rules or concepts.
Scientists then began tracking healthy volunteers for weeks after a single dose to see whether these acute disruptions leave enduring functional or physical marks on the brain. For example, a small 2020 study of 12 healthy volunteers found that some changes in mood and resting brain connectivity were still detectable a month after a single high dose, although others had faded. Similarly, research covered by PsyPost in 2024 found that psilocybin temporarily scrambled brain network connections, with most changes fading within weeks but some, such as reduced connectivity between the hippocampus and a network involved in self-reflection, persisting.
Building on this progression, the current research, led by Taylor Lyons of the Centre for Psychedelic Research at Imperial College London, sought to capture a comprehensive picture of what happens in the brain of someone taking a psychedelic for the very first time. The researchers aimed to connect the immediate, disorganized brain state during a trip to the psychological benefits reported weeks later. They also wanted to test whether a single high dose of psilocybin physically alters the structure of the brain’s white matter.
The study involved 28 healthy adults, averaging 41 years old, who had never taken a psychedelic drug before. Each participant received two oral doses of psilocybin one month apart, always in the same order: first a 1-milligram dose, too low to produce noticeable psychedelic effects, which served as a placebo, and then a 25-milligram high dose known to induce a profound altered state of consciousness. The researchers used this fixed order because they expected the high dose to have lasting effects that could carry over into a later session. To limit expectations, participants were told only that they would receive a variable dose of up to 25 milligrams at each session.
Both dosing sessions took place in a supportive environment featuring low lighting, music, an eye mask, and non-directive guided supervision. During dosing days, the research team recorded electrical brain activity using an electroencephalogram, or EEG, while participants rested with their eyes closed at one, two, and four and a half hours after ingestion. Participants also completed magnetic resonance imaging brain scans before the study began and one month after each dose.
During the high-dose session, the EEG recordings showed large increases in brain entropy, meaning the electrical signals became highly irregular and unpredictable. This peak in disorganized activity occurred one to two hours after taking the 25-milligram pill, aligning with the peak intensity of the psychedelic experience. The recordings also showed a drop in alpha waves, a rhythm of electrical activity common when people are awake and relaxed, while the 1-milligram placebo dose produced no such changes. When asked months later, all but one of the participants who answered (94 percent) rated the high dose as the most unusual state of consciousness of their lives.
When analyzing the brain scans taken a month after the high dose, the researchers found possible changes in the brain’s white matter, the bundles of insulated nerve fibers that carry signals between brain regions. Using a technique called diffusion tensor imaging, which maps how water moves along these fiber pathways, they observed a decrease in water diffusion along tracts connecting the prefrontal cortex to deeper subcortical regions. The prefrontal cortex is involved in complex thinking and decision-making, and this alteration was not observed following the placebo dose.
Other brain imaging measures showed little lasting change. Functional MRI scans, which track blood flow as a proxy for brain activity and which the researchers had designated as their primary outcome, revealed changes one month after the high dose that were largely weak and not statistically significant. One measure, network modularity, reflects how sharply the brain divides into separate, tightly knit networks; it decreased compared with the start of the study but not compared with the month after the placebo dose, and participants with larger decreases tended to show larger gains in well-being.
An earlier analysis of the same participants’ brain scans, covered by PsyPost in 2026, focused on circuits linking the prefrontal cortex with the striatum and thalamus, the same pathways where the white matter changes appeared. Using a different analytic approach, it found that activity in these circuits became more variable a month after the high dose than after the placebo dose, although the difference from the start of the study and the link with improved well-being fell short of statistical significance.
On a psychological level, participants showed improvements after the 25-milligram dose that were not seen after the placebo dose. Their psychological insight scores were about 51 points higher (on a 0-to-100 scale) the day after the high dose than after the placebo dose and remained elevated a month later. Well-being scores rose about 4.7 points more after the high dose than after the placebo dose, on a scale running from 14 to 70 on which the general population averages about 51. Participants also made fewer errors on a computer task measuring cognitive flexibility a month after the high dose, specifically when they had to adapt to hidden rule changes.
When the researchers looked at how these measures related to one another, a pattern emerged. Participants whose brain entropy rose more two hours into the high-dose session tended to report more psychological insight the next day, and greater entropy across the session predicted larger gains in well-being a month later. A statistical analysis suggested that next-day insight partly accounted for the link between entropy and later well-being. Brain entropy did not predict the white matter changes.
“The most surprising – and therefore potentially also important – takeaway is that even in a highly exploratory and multimodal study as described here, psilocybin does not seem to have a pronounced lasting effect on the brain in healthy volunteers,” said Katrin Preller, a junior group leader at the University of Zurich and visiting assistant professor at Yale University who was not involved in the research. “One month after psilocybin administration only very marginal changes were detected. Given the nature of the study, the robustness of these changes needs to be determined in future studies.”
Preller noted that these modest long-term results match an emerging trend in neuroimaging research. “There is accumulating evidence that the neurobiological changes induced by psychedelics may be short lived and alterations are back to baseline levels after a few weeks,” she told PsyPost. “The study therefore confirms these initial reports and extends those by showing that this holds true even for psilocybin-naive participants.”
Martin Korsbak Madsen, a medical doctor and postdoctoral researcher at the Neurobiology Research Unit at Copenhagen University Hospital Rigshospitalet who was not involved in the study, focused on what the findings might reveal about how psychedelics work.
“It is well established that psychedelics can have long-term effects but underlying mechanisms are unknown, and the present study seeks to address this gap by using both EEG and MRI,” Madsen said. “The study confirms previous electrophysiological findings, showing acutely reduced alpha waves and increased complexity. The latter was positively associated with better psychological insight and well-being, so this might be a relevant mechanism for subsequent change.”
Madsen added that the imaging metrics pointed toward potential physiological routes for these mental shifts. “MRI findings suggested that brain function, on average, was less constrained, and that white matter bundles connecting deep and more superficial layers of the brain were also affected,” he told PsyPost. “Taken together the study is a good and thorough first step, and it provides plausible mechanisms, which subsequent studies can test.”
As with all research, there are some caveats to consider. Because all participants received the 1-milligram dose first and the 25-milligram dose second, order effects, such as simply getting used to the testing procedures, could have influenced the results. The authors argue that the changes after the high dose far exceeded any such effects, noting that no changes appeared after the placebo dose. They also acknowledge that their main hypotheses were not registered in advance and that their statistical corrections were applied within, but not across, the many outcomes they examined.
Preller also stressed the study’s exploratory nature. “This is a highly exploratory study which is reflected in a rather limited protection against Type I errors, i.e., the study is optimized for detecting effects but is not rigorously protected against false positive findings,” Preller told PsyPost. “The fixed administration sequence is justifiable in this study but increases the risk that the observed effects are due to habituation or other time-related factors and may not be causally linked to drug administration. All effects observed therefore need replication in better controlled and larger studies before any conclusions can be drawn.”
Madsen also noted that this structure limits how firmly conclusions can be drawn. “A not trivial drawback of the design is the use of a fixed order, meaning everybody got low dose (placebo condition) before the high dose,” Madsen said. “Although this has some benefits, one cannot exclude order effects, and a classic two-arm randomized, double-blind design of sufficient sample size would have been an improvement.”
Interpreting changes in white matter using diffusion imaging can also be complicated. The type of change seen here has been linked to processes such as learning and meditation, but also to nerve fiber injury and aging, so it is unclear what it reflects biologically. The authors speculate it could involve the pruning of weak connections or the growth of new, not-yet-insulated nerve fibers, but they caution against interpreting the finding until it is replicated with more advanced imaging. A second analysis method also failed to detect the white matter changes.
Looking ahead, Preller said studies in patients are needed. “We need more mechanistic long-term studies in patient populations in order to understand whether there is a detectable neurobiological basis for the observed treatment effects,” she told PsyPost. She added that readers should keep in mind that “even the significant findings in this study are highly exploratory and require replication.”
Madsen suggested that future investigations should evaluate psychedelics alongside other established practices that promote personal growth. “Intake of psychedelics is not the only way of improving well-being or gaining psychological insight, and comparing psychedelics to non-drug interventions such as psychotherapy, mindfulness or art therapy, might help elucidate which effects are specific to psychedelics as opposed to general psychological development,” he explained.
The study, “Human brain changes after first psilocybin use,” was authored by T. Lyons, M. Spriggs, L. Kerkelä, F. E. Rosas, L. Roseman, P. A. M. Mediano, C. Timmermann, L. Oestreich, B. A. Pagni, R. J. Zeifman, A. Hampshire, W. Trender, H. M. Douglass, M. Girn, K. Godfrey, H. Kettner, F. Sharif, L. Espasiano, A. Gazzaley, M. B. Wall, D. Erritzoe, D. J. Nutt, and R. L. Carhart-Harris.
-------------------------------------------------
Private, vetted email list for mental health professionals: https://www.clinicians-exchange.org
Unofficial Psychology Today Xitter to toot feed at Psych Today Unofficial Bot @PTUnofficialBot
-------------------------------------------------
#psychology #counseling #socialwork #psychotherapy @psychotherapist @psychotherapists @psychology @socialpsych @socialwork @psychiatry #mentalhealth #psychiatry #healthcare #depression #psychotherapist #psilocybin #psychedelics #brainentropy #wellbeing #psychologicalinsight #cognitiveflexibility #white matter #neuroimaging #NatureCommunications #entropicbrain
-
The Lessons of Direct File
Invest in Capacity Where It’s Hardest to Remove
People working in civic tech should read the excellent paper Public Interest Capacity: The Case of Direct File by Donald Moynihan, Pam Herd, and Xiaoyang Xu, which was just published in Public Administration Review. It’s a comprehensive account of how the IRS built Direct File – a free, high-quality tax filing tool – in-house, and how that tool was eliminated about a year after it launched.
This is an attempt to summarize my primary takeaways from the paper and to think out loud about what I think those in civic tech and working to improve government services should do next.
What Happened to Direct File?
Put simply, while it existed Direct File worked. In its 2024 pilot, 86% of users said it increased their trust in the IRS, and about a quarter of users hadn’t filed the previous year. It was built by a team of roughly 75 people drawn from the US Digital Service, 18F, IRS staff, and outside vendors, working in-house rather than through a traditional government contract.
The paper’s central lesson about the Direct File project is that political will was necessary for it to get started but not sufficient to get it done. The IRS also needed real internal capacity, a skilled team able to build, test, and improve a product rapidly. And it had to withstand the government’s default bias of outsourcing technology, and a tax preparation industry that viewed Direct File as a threat.
With the turnover in Administration, the political will supporting Direct File disappeared. Most of the team, many of them probationary hires with few job protections, left government service. The product itself and the capacity to build and sustain it were lost simultaneously.
Civic Tech’s Bas Toward the Front End
The paper compares Direct File with two other recent digital solution projects – online passport renewal and the rebuilt FAFSA application. All three efforts used a similar approach. Passport renewal and FAFSA survived the turnover in Administration because they don’t have private sector competitors, and because, as one official interviewed for the paper put it, the government has to have a FAFSA product. Direct File was voluntary, so it was easy to terminate.
This underscores an important, but underappreciated aspect of the civic tech movement. Civic tech has tended to favor public-facing products, and I think generally pretty easy to understand why:
- The founding story. Civic tech as we know it today largely grew out of the HealthCare.gov failure, which the public experienced as a broken website. The lessons the movement drew from that experience centered (understandably so) on user experience and product management.
- The talent model. Two- to four-year “tours of duty” traditionally used by groups like 18F and USDS fit a product launch. They are not a good fit for replacing core technology systems, which can take many years and deep knowledge of old code and agency operations.
- The methods. User research certainly can be pointed inward, at the employees who run back-end processes. In practice, however, “the user” usually means a member of the public. That has been civic tech’s focus.
- The funding. Prominent civic tech nonprofits depend on philanthropy, and donors respond to demos, stories, and hard counts of people served. Outside groups also can’t easily impact core government systems, so they tend to build where they can – at the edges.
Public-facing products matter, and they prove what government can do with the right focus and the right resources. But they are visible, sometimes voluntary, and potentially competitive with someone’s business model. Additionally, the knowledge behind them tends to live in small teams that can walk (or be pushed) out the door.
The Case for the Back End
Modernizing core systems has almost the opposite political profile. The IRS’s core systems are non-voluntary: just like FAFSA, they have to work. They serve every taxpayer, including the large majority who file through commercial software or who use a tax preparation service. And better infrastructure can benefit these private providers too, so it can be used to potentially align private interests instead of posing a direct threat.
The rub here is that legacy system modernization has a long, well documented record of failure. The IRS still runs core taxpayer records on systems dating to the 1960s, and decades of large replacement efforts have produced limited results. Those efforts typically ran through the slow, contractor-led, all-at-once approach the paper calls out.
Working on back-end systems also tends to be invisible, which can make it hard to fund. It depends on long-term stewardship rather than short tours of duty. For most of the time that civic tech has existed in the wake of HealthCare.gov, a new public-facing product might be been the only way to show real results within a finite political window. But I think that calculation is changing.
AI Changes the Game
AI coding tools have changed software development dramatically in just the few years since Direct File started. They change the economics of legacy modernization in ways that weren’t available when Direct File was chosen as a flagship project.
The SpecOps Method is a good example of this. Rather than having AI translate old code directly into new code, it uses AI to turn legacy systems into plain-language specifications. Domain experts can then verify those specifications, and the verified versions become the authoritative definition of how the system should behave. Modern code is generated from them, and components are replaced gradually rather than all at once.
This approach might be a way to address the weaknesses the paper identifies:
- Knowledge outlasts the team. Direct File’s biggest loss was the people who understood it. A verified, version-controlled specification survives staff turnover in a way a team’s expertise doesn’t.
- The key people are program experts. Specifications are reviewed by policy and program staff, not programmers or engineers. That makes longer-tenured career employees rather than a rotating crew of technologists the arbiters of truth for system behavior. Additionally, the technical teams needed for this work are likely small.
- Gradual replacement instead of big bangs. Replacing components piece by piece answers the main reason past IRS modernization efforts have failed in the past.
- Outside groups can help too. Reusable AI instruction sets, such as tools for understanding COBOL, can be shared across agencies. That is work civic tech organizations can build, maintain, and show to funders without needing access to core systems.
The open source source code for Direct File could also give this work a head start. Its Fact Graph encoded tax law as machine-readable rules that IRS experts have already reviewed. In essence, that makes it a verified reference for what the system is supposed to do. AI tools could compare legacy behavior against it and surface only the discrepancies, which would more efficiently allocate the time of experts in reviewing the system. With the caveats that the Fact Graph covers filing rather than downstream processing, and that it probably needs updating for recent tax law changes, it might actually turn out to be Direct File’s most lasting contribution.
Industry Access Must be Earned, and Enforced
Earlier I said modern backend infrastructure could align private interests with the goal of easier tax filing. Given the history laid out in the Direct File paper, I don’t think this can’t be left solemnly to goodwill. Under the old Free File partnership, the government ceded its ability to build its own tool in exchange for an industry promise to reach 70% of taxpayers. Between 2015 and 2020, only 1.6% to 2.7% of tax filers used it, while companies steered people toward upsells and paid products.
Any new arrangement with industry has to reverse that outcome. Modern backend infrastructure would give the IRS something valuable to offer, particularly access to the wage and income data it already holds so tax returns could be prefilled. Access to streamlined systems and richer data could be conditioned on:
- Honest marketing, with no “free” products built to push users into paid upgrades
- A genuinely free tier for simpler tax returns
- Limits on using tax data for upselling
- Letting users take their return data to another provider or a public tool
The government must absolutely keep its leverage. There should be no limits on building public tools if it makes sense in the future. It should insist on measurable standards with real audits, and access to IRS systems that can be suspended if industry doesn’t uphold its end of the bargain.
Interestingly, Direct File itself can be part of that leverage. This effort proved that a government team can build a filing tool people genuinely like, and the work behind it still exists. If industry doesn’t hold up its end, rebuilding a public option is a realistic fallback, not some empty threat. The devil is in the details here, and much would need to be worked out. Still, I think this could work.
Where to invest in capacity
While we’re on details that need to be worked out, there are certainly a few. Verification of system specifications depends on exactly the kind IRS staff who have been leaving recently. Specifications derived from legacy code might expose places where the system’s behavior differs from the law or regulation, and resolving those gaps is sometimes a political or legal decision.
The discipline of keeping specifications current has to survive changes of Administration. And back-end work still produces little that is visible to the outside, so it will need to be funded as risk reduction and knowledge preservation rather than as a shiny new public-facing service.
Even so, the approach I think could make sense looks something like this:
- Specify and modernize the core IRS systems and tax rules that everyone depends on.
- Expose that infrastructure through modern interfaces, available to commercial software on enforceable consumer-protection terms.
- Build public tools as thin layers on top, cheap to create and easy to rebuild at some point if they are eliminated.
The Direct File demonstrated that government can build the kind of technology solutions that people want and need. Its abrupt elimination also showed that a visible product that competes with private sector offerings is the easiest kind of capacity to remove. The lesson for those who will do this work into the future is to invest in capacity that is hard to remove.
#AI #artificialIntelligence #ChatGPT #government #llm #tax #technology -
Lakers’ Walker Kessler makes preseason debut in against Warriors https://www.rawchili.com/nba/858361/ #Basketball #california #GoldenState #GoldenStateWarriors #GoldenState #GoldenStateWarriors #LasVegas #LosAngelesLakers #LukaDončić #NBA #Sports #WalkerKessler #Warriors
-
Lakers Forward Sends Message About Last Season’s Struggles https://www.rawchili.com/nba/858357/ #Basketball #LA #LALakers #Lakers #LALakers #LosAngeles #LosAngelesLakers #LosAngeles #LosAngelesLakers #NBA
-
Lakers Injury Report: Walker Kessler active vs Warriors, Luka Dončić out https://www.rawchili.com/nba/858285/ #Basketball #LA #LALakers #Lakers #LakersInjuryReport #LALakers #LosAngeles #LosAngelesLakers #LosAngeles #LosAngelesLakers #NBA
-
Lakers Pushed to Trade for Young 6’10” Double-Double Gem Amid Kessler Injury https://www.rawchili.com/nba/858212/ #Basketball #LA #LALakers #Lakers #LALakers #LosAngeles #LosAngelesLakers #LosAngeles #LosAngelesLakers #NBA
-
Lakers Could Trade Austin Reaves For Finals MVP Superstar Wing: Rumor https://www.rawchili.com/nba/857923/ #Basketball #LA #LALakers #Lakers #LALakers #LosAngeles #LosAngelesLakers #LosAngeles #LosAngelesLakers #NBA
-
Lakers insider makes Walker Kessler availability clear ahead of LA debut https://www.rawchili.com/nba/857785/ #Basketball #LA #LALakers #Lakers #LALakers #LosAngeles #LosAngelesLakers #LosAngeles #LosAngelesLakers #NBA