Podcasts by Dirk Werner

The German episode calls this the emotional fog: a great stretch is almost entirely forgotten the moment stress sets in, while a crisis feels as though it had always been there and would never end. What is missing in that moment is not a sudden insight but a reliable record saying: yesterday you were doing fine. The difference between “the day was bad” and “mood is low, but sleep and energy were fine” is exactly what keeps one bad meeting from becoming a verdict on your whole life.

The conversation uses a car dashboard: one warning light only tells you something is wrong and leaves you stranded on the motorway. Only when you see eight areas separately — mood, sleep, energy, closeness, body, focus, calm and meaning — can you tell whether the engine is running and merely low on oil. Part of that is the absence of any completeness requirement: two of eight values on a hard day are enough, and a skipped day breaks no streak.

The episode names three: begin an unpleasant task for exactly five minutes, with no obligation to finish it. Spend one evening without the automatic reach for your phone. Accept a friendly piece of feedback without instantly talking it down. Afterwards you look at it plainly: what did I expect, what actually happened? There is no pass and no fail — even five minutes you abandoned are a data point. And if the tension is too high for that, calming comes first: a three-minute breathing pause, grounding, progressive muscle relaxation or the inner safe place. The episode puts it in one line: nobody solves a sudoku on a rollercoaster.

The episode explains why that is down to how ordinary trackers are built, not to you: a single grade for the day hides the pattern instead of showing it. It also fits if you want to know why an AI here deliberately interprets nothing, advises nothing and never diagnoses. The closing thought needs no app at all: for one day, in conversation, genuinely just listen, summarise briefly and ask honest questions — instead of reflexively repairing the other person's problem.

The episode traces it back to the name: “Sundara” is Sanskrit for beautiful or harmonious, “sunda” is Japanese for having become clear. The idea behind it: your mind does not have to be flawless or optimised in order to be clear — you can be messy and clear about it at the same time. That is why Sunda is defined by what she refuses to do: no interpreting, never a diagnosis, no life hacks. The episode compares her to the mirror in a dance studio: it does not critique your routine or tell you where to put your feet, it reflects your posture so you can correct it yourself.

The episode names the reflex: we treat our own psychology like a buggy operating system — find the broken part, apply a patch, track the improvement. Against that it sets one line from the documentation: a bad Tuesday feels like a bad week. Eight separate dimensions therefore work like an equaliser rather than a single master volume: closeness and meaning may have been perfectly fine today, and sleep alone sits at zero, dragging your perception of the whole day down with it. And when the pattern engine notes that focus tends to drop three days after sleep and closeness fall together, that is a hypothesis and not a truth — correlation is not causation, and the episode says so explicitly.

The episode's example: for the next three days, whenever a message makes your blood boil, set a timer and wait exactly five minutes before replying. Afterwards you answer two questions — what did I expect, and what actually happened? In the episode's own example the answer is: I expected that delaying would feel like losing the argument. What happened was that the message was not an attack at all, just badly phrased — and I saved myself a three-hour fight. There is no pass mark and no streak to break: a missed day is simply a day you did not need to journal.

The episode is unusually candid here: independent third-party security audits are still pending, it is not fully public which AI services process the text, and there are no peer-reviewed studies on long-term effect. It is just as clear about what is already in place: a hard-coded filter every generated sentence has to pass; crisis language that switches the mirror off and surfaces emergency numbers instead; and a strict separation between the developer's therapy practice and the app — users are explicitly not his patients.

It also goes further on two things the German episode leaves aside: that correlation is not causation, and where the app's limits currently are. Listening to both gives you two angles — not the same thing twice.

The German episode sets this image against the idea of a bang: while driving you barely notice, you steer against it a little, you compensate — until one day you are scraping along on the rim. What is missing is the voice from the satnav: take a wrong turn in a car and it says “please turn around”. Drift away from your own convictions and nobody says anything. And the ending turns it around once more: self-doubt and adapting are not a defect but a program written in childhood, when conformity secured belonging. It still runs flawlessly — except that you now hold the administrator rights to rewrite it line by line.

The episode calls it the yes-no list: write down every commitment you made in a week and mark honestly which came from conviction and which only avoided a conflict. Alongside it stand the signs that appear earlier — diffuse restlessness, irritability, and the feeling of being merely a passenger in your own life. The episode calls this state cognitive dissonance: holding two contradictory realities at once costs energy continuously, even when nothing visible is happening.

The three-minute inventory asks: what am I feeling? Where do I notice it in my body? What set it off? The hard part is the rule behind it: record perceptions only, no judgements. So “my neck has been tight since that email from my boss” instead of “I am badly organised and will never manage this”. Then ten unvarnished lines on three evenings, unedited and without a single word crossed out — whoever crosses out is already negotiating with outside expectations again. Then the weekly review in three questions. And the one question to three people from different parts of your life: when do I seem most like myself to you?

The episode calls forced positivity one of the most dangerous pieces of pop psychology: reframing difficult circumstances against your own perception deepens the dissonance and suppresses your own warning signals. The counter-model is realistic optimism — not denying the pain, but being able to notice what went well again. It also fits a stretch where you keep saying yes and afterwards could not say what for.

The episode is blunt about the opposite: a rigid persona that refuses to adapt is a defence mechanism, not authenticity. What actually wears you down is the slow sacrifice of your own needs to an environment, until intrinsic motivation decays and things that once came naturally stop giving you anything. And it names the cost of the popular remedy: forcing an optimism you do not feel consumes so much energy that it drains you faster than pessimism would.

The episode compares them to pain receptors: touch a hot stove and the pain says nothing about your character, only move your hand. Both mistakes are available — deadening the nerve, or panicking and knocking over the pot of boiling water. The point is to read the signal and move deliberately. In practice the body reports first: chronic restlessness, irritability, and the sense of being a guest in your own life. The yes-no list makes it countable — the episode calls it an audit of your bank statement, except that what you see is not where your money goes but where your identity is being spent.

Tarek, a consultant, was arguing positions at work he no longer believed. He did not quit; he negotiated a four-day week and spent Tuesdays volunteering as a debt counsellor. It took eighty Tuesdays before the career change was real. Hannah lost her husband and her library job within fourteen months and set three non-negotiable weekly commitments: swimming on Mondays, calling her sister on Wednesdays, groceries for a neighbour on Fridays. None of it addressed the grief directly. It built a frame in which grief could happen without filling everything.

Miriam runs a hospital emergency department; after a shift in which two resuscitations failed she sits in her car at four in the morning and cannot turn the key. She does not quit. She writes one sentence per shift in a ring binder: what was beyond my control today. After six months and about a hundred sentences the healthcare system is unchanged — but the line between her actual responsibility and everything else has been drawn. The episode ends somewhere unexpected: with an apprentice watching an older colleague quietly fix a structural error in a roof truss that nobody would ever have seen — and with the question of who is watching you.

This one takes its time with two other case studies — Tarek's eighty Tuesdays and Hannah's three weekly commitments — reads emotions as pain receptors rather than noise, and closes with the apprentice on the building site. Listening to both gives you two angles, not the same thing twice.

The bread is the occasion, not the subject. Judging someone's character over a forgotten errand means reacting to something older than the errand.

For Jung this is neither a conscious trick nor a character flaw but something involuntary: you do not create a projection, you find it already formed in your own perception. It is strongest where the shadow lies, in everything we refuse to be ourselves.

The term moves the question away from one person's fault and toward what two people produce together. That is why arguments in which both sides present accurate evidence so rarely resolve anything.

What is new is not the projecting itself but its content: no longer a private ideal but a social, algorithmically amplified super-ideal. Filtered photos and conflict-free feeds supply the image against which the real person at the table is then measured.

The partner gradually begins to live it out, so that the projection ends up confirmed by reality. Jürg Willi described the same interplay in couples under the name collusion.

He really is carrying more of the load; she really is living with someone who checks and reminds and takes over. Neither sees how that reality came about, so each keeps producing evidence that holds up.

As long as we love, we leave them unfinished and able to surprise us. And the image acts back on them: we hold it up until they begin to match it.

So the polished ideal stays on the screen, and the sweating, stuttering real person across the table is measured against it.

Someone convinced they are not lovable already holds a file of counterevidence. The positive sentence does not erase it, it calls it up - and makes it more present and more painful in that moment than before.

No inner contradiction arises, because nothing is asserted that runs against your own experience - you steady a different pillar instead of defending the one under attack.

On a twelve-item scale the induction added roughly one and a half further responses on average. What remains once trance is removed is a surprising amount - it simply needs no open door.

Crediting the app, the voice on the recording or a special trance leaves you with something pleasant and little learned about yourself. The episode puts this attribution at its centre; whether it makes a lasting difference is a hypothesis.

The claim does not overwrite the record, it retrieves it - which is why people with low self-esteem felt worse than a control group that said nothing at all.

Nothing about the message had changed. What had changed was how much they needed to defend themselves against it.

Sit still long enough and the arm does press against the armrest; attention finds the sensation that is already there. The sentence gets confirmed, and a confirmed sentence is something “I am calm” never becomes in a moment of anxiety.

Half were then told the truth and had to credit themselves. In the next round those men tolerated considerably more. The episode treats this as central; the essay marks the lasting benefit as a hypothesis.

Fluent, confident language acts as a signal of truth, even though it says little about the answer’s accuracy.

In experiments, participants’ accuracy rates largely mirrored the accuracy of the AI.

The problem lies less in outsourcing the work than in outsourcing control.

It’s also helpful to decide in advance which topics always require verification.

Fluent, confident language acts as a signal of truth, even though it says little about the answer’s accuracy.

In experiments, participants’ accuracy rates largely mirrored the accuracy of the AI.

The problem lies less in outsourcing the work than in outsourcing control.

It’s also helpful to decide in advance which topics always require verification.

The eighteenth, Gradual Normalization, is already under way according to the conversation.

The episode cites the 1975 study by Wagenaar and Sagaria and METR's measurements, which put the doubling time of the task length an AI can handle at roughly 90 to 131 days.

The episode shows this with the bill from Sanders and Casar and Trump's line that whoever wins in AI wins it all.

According to the episode a GPS is harmless as long as paper maps exist and someone can read them. The danger is not handing a task over but being unable to return.

Only about 1 percent opened the sources cited in the summary. The episode reads this as a first stage in which we hand over checking the source.

The eighteenth, Gradual Normalization, is already under way according to the conversation.

The episode cites the 1975 study by Wagenaar and Sagaria and METR's measurements, which put the doubling time of the task length an AI can handle at roughly 90 to 131 days.

The episode shows this with the bill from Sanders and Casar and Trump's line that whoever wins in AI wins it all.

According to the episode a GPS is harmless as long as paper maps exist and someone can read them. The danger is not handing a task over but being unable to return.

Only about 1 percent opened the sources cited in the summary. The episode reads this as a first stage in which we hand over checking the source.